1

4 SELECT COMMITTEE TO INVESTIGATE THE

5 JANUARY 6TH ATTACK ON THE U.S. CAPITOL,

6 U.S. HOUSE OF REPRESENTATIVES,

7 WASHINGTON, D.C.

10

11 DEPOSITION OF: J. SMITH

12

13

14

15 Monday, May 9, 2022

16

17 Washington, D.C.

18

19

20 The deposition in the above matter was held via Webex, commencing at 11:14

21 a.m.

22 Present: Representative Schiff.

 2

2 Appearances:

5 For the SELECT COMMITTEE TO INVESTIGATE

6 THE JANUARY 6TH ATTACK ON THE U.S. CAPITOL:

8 , SENIOR ADMINISTRATIVE ASSISTANT

9 PROFESSIONAL STAFF MEMBER

10 STAFF ASSOCIATE

11 , INVESTIGATIVE COUNSEL

12 , INVESTIGATIVE ANALYST

13 CHIEF CLERK

14 COUNSEL

15 SENIOR COUNSEL & SENIOR ADVISOR

16

17

18 For the WITNESS:

19

20 ALEXIS H. RONICKHER, ESQ.

21 Katz, Marshall & Banks, LLC

22 1718 Connecticut Avenue NW, Seventh Floor

23 Washington, D.C. 20009

 3

2 - Good morning. This is the deposition of J. Smith, conducted by the

3 United States House of Representatives Select Committee to Investigate the January 6th

4 Attack on the United States Capitol pursuant to House Resolution 503.

5 This will be a staff-led deposition, although members may of course choose to ask

6 questions.

7 At this time, I'd like to note for the record that during this deposition staff will

8 refer to the deponent as "J. Smith" in order to protect their identity. Any references to

9 the witness's legal name will be redacted from the record.

10 Do you agree to that setup?

11 Ms. Ronickher. You have to unmute and answer on the record.

12 The Witness. Yes, I do.

13 - Thankyou.

14 My name i s - and I'm an investigative counsel with the select

15 committee. With me today from the select committee staff are

16 investigative analyst; counsel; senior counsel and

17 senior advisor; professional staff member; chief clerk.

18 I will announce any additional participants as they join, including any members

19 that may join. And as I said, this will be a staff-led interview, but if any members do join,

20 they may of course ask questions.

21 In a few moments, I'm going to hand it over to - t o begin questioning,

22 although I will also be asking questions to you throughout the deposition. There may be

23 other staff members, such as who will ask questions as

24 well. If you don't understand a question, simply ask us to repeat it.

25 J. Smith, you have been subpoenaed by the select committee to compel your
 4

1 presence at this deposition. Can you confirm to me now that you have received a

2 subpoena from this select committee for this testimony today?

3 The Witness. Yes, I can.

4 - Thank you. And do you understand that you are appearing here

5 today pursuant to the subpoena?

6 The Witness. Yes, I do.

7 - Thank you.

8 Because this is a deposition, you may refuse to answer a question only to preserve

9 a privilege that is recognized by the select committee. If you refuse to answer a

10 question based on one of these privileges, we may either proceed with the deposition or

11 seek a ruling from the chairman on the objection. If the chairman overrules such an

12 objection, you are required to answer the question.

13 My goal today, and all of our goals today, is to ask questions relevant to the select

14 committee's investigation with the hope that you will answer. If you have an objection

15 or a privilege assertion, we'll ask that you or your counsel assert it for the record. Of

16 course, we don't anticipate having that problem today, and we thank you for your

17 cooperation.

18 In the event that you or your counsel do raise any objections, I will seek to clarify

19 the basis for each objection. Ultimately, the more detail you can provide on each

20 objection, the easier it will be for the select committee to consider the objection in full.

21 There is an official reporter transcribing the record of this interview, so please

22 wait until each question is completed before you begin your response, and we'll try to

23 wait until your responses are complete before we ask another question.

24 Additionally, the stenographer cannot record any nonverbal responses, like

25 shaking your head or nodding your head, so it is important that you answer each question
 5

1 with an audible, verbal response. And, of course, if you forget to do that, we may

2 prompt you to do so or note for the record that you were nodding your head.

3 We also ask that you provide complete answers based on the best of your

4 recollection. If the question is not clear, just please ask for a clarification. And if you

5 don't know the answer or know only certain details of an answer, simply tell us.

6 Similarly, if there's anything that you need to discuss with your attorney at any

7 point in private, please let us know, and we'll take a break so you can have that

8 conversation off the record. Additionally, I'll plan to take a break around every hour,

9 but if you need breaks at a more frequent interval for comfort or for conferring with your

10 attorney, just say so.

11 This interview is under oath, and you are obligated to tell the truth, under Federal

12 law, as if you were speaking to the DOJ or the FBI. It's unlawful to deliberately provide

13 false information to Congress. For this interview, providing false information could

14 result in criminal penalties for perjury and/or false statements. And this is something

15 that we tell all of our witnesses.

16 You are not obligated to keep the fact of this deposition or what we discuss

17 confidential. However, we are obligated, under select committee rules and House rules,

18 to do so. You are free to tell whomever you wish that you met with us.

19 And, again, please let us know if you need any breaks or would like to discuss

20 anything with us or with your attorney off the record.

21 Because this deposition is under oath, right now I'll ask you to raise your right

22 hand to be sworn in by the reporter.

23 Ms. Ronickher. And you should unmute.

24 The Reporter. Do you solemnly declare and affirm under the penalty of perjury

25 that the testimony you are about to give will be the truth, the whole truth, and nothing
 6

1 but the truth?

2 The Witness. Yes, I do.

3 - Thank you.

4 At this point, I want to confirm with the witness and counsel that they are okay

5 with-conducting questions, given that he is not an lawyer and is an

6 investigative analyst with the select committee.

7 Ms. Ronickher. Yes.

8 The Witness. Yes as well.

9 - Thank you.

10 So, without objection to that, I will hand it over now t o - t o begin with

11 some background questions.

12 Thank you so m u c h , _

13 EXAMINATION

14 BY
15 Q J. Smith, I was wondering if you could tell us about your highest level of

16 education, just to establish identity and background.

17 A Yes. My highest level of education is a Juris Doctorate degree.

18 Q Excellent. Thank you. I'm trying to keep these questions at a certain level

19 of vagueness, as well, to protect your identity.

20 What was your role while you were employed at Twitter?

21 A I was a senior safety policy domain specialist.

22 Q Approximately how long did you work in this role?

23 A Almost 2 years.

24 Q Were you employed at Twitter during January 6, 2021?

25 A Yes, I was.
 7

1 Q What were your responsibilities in your role at that time?

2 A My team was responsible for writing and enforcing certain aspects of

3 Twitter's rules.

4 Q And to whom did you -- well, we can strike that question. I understand

5 that's not really germane.

6 I wanted to start off with exhibit 1, which is a policy document we obtained from

7 Twitter on coded incitement to violence. I thought we could use this to foreground our

8 conversation.

9 Can we pull that up?

10 Thank you.

11 Can you everyone see this okay?

12 A Yes.

13 Q Excellent. So do you recognize this document?

14 A Do you mind scrolling down a little bit so I could take a full look at it?

15 Yes, I do recognize this document.

16 Q For the record, do you think that you could summarize the policy this

17 document spells out?

18 A Yes. This document is specific guidance related to a policy called "coded

19 incitement to violence" that the safety policy team worked on.

20 Q What led to the development of this policy, which looks like it was last

21 updated on November 4, 2020?

22 A Yes. What led to the development of this policy. We were asked

23 by -- Alexis, this is a moment where lawyers -- we were asked by our lawyers to work on

24 policies that would encapsulate some of the language that we had seen coming from the

25 former President, especially coming out of one of the debates.

 8

1 Q Was that language specifically related to "stand back and stand by" or any of

2 the other phrases identified?

3 A Yes. It was specifically related to the President's use of the phrase "stand

4 back and stand by" in relation to the Proud Boys and other extremist groups.

5 Q If we could scroll down a little in the document.

6 So are you able to tell us who the individual was that signoff was required from to

7 apply this guidance to any user?

8 A I would believe -- my recollection would say that it would either be Del

9 Harvey or Vijaya Gadde or Jack Dorsey would be the three individuals who would be able

10 to sign off on those high-level decisions.

11 Q Okay.

12 For how long before this policy was written were dog whistles or other

13 incitements to violence to extremist groups by the President a concern for your team and

14 your leadership at Twitter?

15 A They had been a concern from several -- from several members for my entire

16 time at Twitter. But it was not, again, until the debate where the former President used

17 the specific language of "stand back, stand by" where it became a directive for us to

18 determine how to handle those sorts of phrases and that type of specific language on the

19 Twitter platform.

20 Q And during previous incidents where the President might have done this,

21 were there times where members of your team advocated for a similar policy before the

22 debate but were rebuffed or it just wasn't escalated?

23 A I do not believe that we advocated for a similar type of policy. In instances

24 prior to this, our team would write policy assessments based on the policies that we

25 already had on the books in order to interpret how we should, again, apply Twitter's
 9

1 policies to the language of the former President.

2 Q So my understanding is that this policy was escalation-only. Do you know

3 why this was an escalation-only policy?

4 A Yes. This policy was an extremely nuanced policy, and because of the

5 nuance that was required and the additional context that was required into making

6 recommendations for decisionmakers, it was determined that the individuals who should

7 be doing that assessment should be on the safety policy team. And so this material was

8 escalated to the safety policy team for review.

9 Q Were strikes issued to accounts which violated this policy?

10 A I do not remember. It might be -- within the document there should be a

11 recommendation of how strikes should be applied, but, from my memory, I believe the

12 recommendation was for an abuse one-off, which would not have applied a strike.

13 Q To your recollection, did that change over time as the policy evolved? Did

14 the application of strikes change?

15 A Yes. There was a time on January 6th itself in which there was a decision

16 made not to just issue a strike but actually to permanently suspend, which is the highest

17 level of punishment for infraction or enforcement against Twitter's rules based upon the

18 coded incitement to violence policy.

19 Q And when was that decision? I'm sorry? You said --

20 A January 6th. The morning of January 6th.

21 Q January 6th. And that would have been the afternoon East Coast time? Is

22 that right? Morning Pacific time?

23 A Morning East Coast time.

24 Q Morning East Coast time. Okay.

25 I just want to confirm -- if we scroll to page 9 of this document, I want to verbally

 10

1 confirm that "content that meets the assessment criteria above will be actionable as

2 abuse one-off, no strikes."

3 It's at the very bottom of the document -- or, of page 9.

4 A Yes. Yes, that is what I was referring to with my memory, again, saying

5 that, underneath the coded incitement to violent policy, our recommendation was to not

6 issue what we called a strike. And that was specifically because this policy was not

7 public-facing or public-known, and we did not want to be issuing strikes to users who

8 were violating a policy that they did not know existed.

9 Q Okay. Thank you very much.

10 I believe - a s some followup questions before we move to the next

11 exhibit.

12

13

14

15
-

Q
BY- Yes, thank you,

So I wanted to zoom out and set a couple of broader baseline

16 understandings of this policy and the origin of this policy.

17 Thank you, for scrolling up to the top.

18 So can you give me a better sense of the precise timeline for when this policy was

19 first brainstormed and when it was implemented?

20 A Yes --

21 Ms. Ronickher. And, J. Smith, if you had conversations with any Twitter counsel,

22 you can reference that you had a conversation with Twitter counsel, but do not state

23 what the content of that conversation is.

24 The Witness. Thank you.

25 do you mind repeating the question? You asked me two parts, and I
 11

1 want to make sure that I answer them.

2 BY-

3 Q Sure. So the first part is: Can you give me a sense of the general time line,

4 when this policy was proposed? And, sort of, then walk us through the timeline of

5 implementation.

6 A Yes.

7 So, then, the proposal for the creation of this policy, I believe, if my memory

8 serves correct, the debate in question happened at the -- in October, I believe.

9 And so, after that election -- or, not the election; excuse me -- after that debate,

10 again, we spoke with counsel, and, following that conversation, members of the safety

11 policy team began writing this coded incitement to violence policy during, I want to say,

12 November. The beginning of November was when the policy was created and written

13 for the most part.

14 Q And you say created and written early November.

15 A Yes.

16 Q Is that the same time at which it went into effect?

17 A No. The question around this policy going into effect is a big one. It

18 was -- our team continually struggled and asked for clarification about when and if this

19 policy was going into effect and how. And -- yes.

20 Q Thank you, J. Smith. That's very helpful.

21 And I don't want to jump ahead too much, but would you say that lack of clarity

22 on enforcement continued throughout the post-election period?

23 A Absolutely.

24 Q Thank you.

25 So, first, for the record, I wanted to make sure, this is the first debate you're
 12

1 referring to, where President Trump made the comment about the Proud Boys standing

2 back and standing by?

3 A I believe so. I'm not exactly sure which debate it was, but it was one of the

4 debates, yes.

5 Q Okay. Yeah, well, I believe that was the very end of September, which

6 would make sense -- great.

7 The witness gave a thumbs-up.

8 A That sounds right, yes. That sounds about right.

9 Ms. Ronickher. J. Smith, just make sure that you answer verbally.

10 The Witness. Yes.

11 BY

12 Q And so I also wanted to ask another broader question about the role of the

13 safety policy team. You mentioned that team a few times in the context of drafting and

14 formulating this policy. So what was the general role of that team at Twitter? Why

15 was it involved in the creation of this policy?

16 A Yes. So the safety policy team was, again, responsible for the development

17 and enforcement of Twitter's rules, specifically around safety and privacy. And so that

18 specifically included policies around abuse, harassment, hateful conduct, violence,

19 privacy, sensitive media, suicide, self-harm, and at times misinformation.

20 And so, with that purview, our team, again, was responsible for writing the

21 external policies that you see on the website called "The Twitter Rules," as well as the

22 internal guidelines for how to understand and distill those rules internally to all of the

23 content moderators around the world. And because one of the purviews of the safety

24 policy team was, again, violence or incitement, this content fell underneath our purview.

25 Q Thank you. That's very helpful, J. Smith.

 13

1 And I wanted to follow up on that to ask if there had been conversations within

2 your team about the need to have a policy that captured some of these concerns prior to

3 the end of September, early October, to the point with the debate.

4 A I do not believe that there were any official conversations about the creation

5 of a specific policy in order to deal with any of this content.

6 Q Thank you. That's really helpful.

7 And, in that case, can you kind of walk us through why the first debate was such a

8 transformative moment in terms of how the safety policy team and Twitter, it sounds like,

9 more broadly viewed their current policies as inadequate?

10 A Yes. Actually, could you repeat that question for me?

11 Q Sure. I guess, could you explain why the first debate and the former

12 President's comments at the first debate precipitated the need for -- or the perceived

13 need for this policy?

14 A Yes.

15 Leading up until that debate, we had seen language on the service being used by

16 the former President that was increasingly bumping against Twitter's policies and

17 Twitter's rules. Up until that point, we had begun using the public interest interstitial

18 more -- well, we had begun using it more frequently on the President's account -- former

19 President's account. And we used that intervention when content was found to be in

20 violation.

21 And so, seeing the trajectory of the language becoming increasingly more towards

22 the violative side or the prohibited side of the platform, it was recognized that the

23 language within the debate that we have been speaking about was too far.

24 And there was a concern -- there was a concern that existed at that time around

25 anything that the former President saying in a public forum, him then also saying on the
 14

1 Twitter platform, because there was a correlation between those things.

2 And so, seeing the former President and hearing the former President using that

3 language that was extremely coded, there was a concern that that language would then

4 make its way onto the service and that our current policies, especially around

5 incitement -- because our policies around incitement were actually in flux at the

6 time -- would not be able to address that type of coded language.

7 And there was a concern about the implication and the impact that that type of

8 coded language would have, not just on the service but, specifically, offline harm that

9 could occur because of that language.

10 Q Thank you.

11 I have a couple of questions on that. First, how did Twitter, internal

12 conversations at Twitter, react to the President's statement at the debate? We've

13 talked a lot about, sort of, in policy terms, the implications of what he said, but what did

14 you actually think or fear he was saying?

15 A Myself or Twitter?

16 Q I guess you could say both for yourself at the time and to the extent you're

17 aware of other folks that -- or had conversations with other folks at Twitter about it.

18 A I'm going to ask you to repeat the question one more time again.

19 Q Sure. So I guess, first, what were you concerned about when the President

20 made those remarks at the debate? And, second, did you have conversations with

21 others at Twitter, and what were their concerns?

22 A Yes. My concern was that the former President, for seemingly the first

23 time, was speaking directly to extremist organizations and giving them directives. We

24 had not seen that sort of direct communication before, and that concerned me.

25 As far as Twitter is concerned, again, within a matter of days, members of my

 15

1 team and members of the legal team responsible for enforcement decisions were in

2 conversation about making sure that we were able to handle those types of language and

3 phrases and content if it were to arrive to our site. Because there was a fear that we

4 would not be prepared and we would not have a policy in place to be able to enforce

5 against that content and that it would lead to, again, sort of real-world harm.

6 Q Thank you.

7 So, just to clarify further, you were worried and others at Twitter were worried

8 that the President might use your platform to speak directly to folks who might be incited

9 to violence?

10 A Yes.

11 Q And the policy was a direct response to that fear?

12 A Yes.

13 Q I just have one more question, and we have to move on. We could go all

14 day on one exhibit.

15 Can you explain to us what the public interest interstitial was or is?

16 A Yes. So the public interest interstitial is one of the first -- one of the first

17 type of enforcement decisions of its kind which created a label, or a tombstone, in

18 instances in which a policy was found to be -- or, excuse me, a piece of content or tweet

19 was found to be in violation.

20 There were certain criteria, whether that person -- and these are all public; all the

21 criteria is public. If an individual met that specific criteria, their information and content

22 was deemed to be what was called "within the public interest." And because it was

23 within the public interest, rather than removing the content, the content was then

24 labeled with information that said -- language that said that the tweet had broken

25 whatever policy and that it was against the rules.

 16

1 That label also -- it stopped all engagement with that content so the content could

2 not be retweeted, it could not be quote-tweeted, it could not be liked. The tweet

3 essentially, again, was a dead tombstone.

4 Q That's really helpful.

5 And so, in that context, it was the belief that the potential harm that could come

6 from direct appeals to the public by the former President for potential incitement could

7 not simply be -- could not simply be countered with the tombstone message anymore?

8 Yeah, I could rephrase the question.

9 A Please. Thank you.

10 Q Yeah. I was thinking about it as I was saying it.

11 Is it fair to say that those public interest interstitial messages were no longer seen

12 as adequate once Twitter became concerned about the President making direct appeals

13 to incitement on the platform?

14 A I don't know if that was clear, but I believe it was a concern.

15 Q Great. Thank you.

16 Well, thank you for bearing with me, and I'll hand it back to

17 Thank y o u , _

18 And thank you, J. Smith.

19 BY
20 Q If we could, I'd like it move on to exhibit 2.

21 Great. Thank you.

22 So this document is the "Post-Election Protests and Calls for Intervention."

23 Our understanding is that on November 5th, just 1 day after the last update to the

24 coded incitement to violence policy and just 2 days after the 2020 Presidential election,

25 the coded incitement to violence policy was integrated into this larger policy guidance.
 17

1 First off, is that correct? And second off, do you know the rationale for that

2 decision?

3 A I am still a little bit unclear as to how -- I was not involved in the

4 conversations and I'm a little bit unclear as to how the coded incitement to violence

5 policy was folded or included within this post-election protest and calls for interference

6 guidance. I do believe that your understanding that that somehow happened -- I

7 believe -- I do remember hearing that.

8 And, again, it was a continued question from myself and my team, what the

9 details of that encompassed and how, again, the policy was exactly incorporated into this

10 one.

11 Q So was this a document your team would have worked with or worked on

12 closely?

13 We can scroll down a little, if that would help.

14 A We would have contributed to this document, yes, but it would not have

15 been solely written by the safety policy team.

16 As you can see, it says, "This document sets out a coordinated approach by health

17 policy teams." So the health policy teams, which was several more teams -- safety

18 policy was included within health policy, but there were several other teams that were

19 included within health policy that would've contributed to this document.

20 Q So "health policy" refers more to the health of the overall site, not public

21 health?

22 A Yes and no. So health policy is a specific designation for the departments

23 that include what is called "site integrity" as well as "site policy." Safety policy sat

24 underneath the site policy umbrella.

25 Q Who would have had oversight over the crafting of this document? Who
 18

1 sat at the junction of all those teams?

2 A Del Harvey --

3 Q Del Harvey.

4 A -- who was the vice president -- or, was -- excuse me -- the former vice

5 president of trust and safety --

6 Q Thank you.

7 A -- at Twitter.

8 Q Yeah. Yes.

9 Could you, for the record, just to the best of your recollection, describe what this

10 policy was for, what it covered, how it was applied?

11 A Yes.

12 Q Take your time.

13 A Would you mind repeating the question again, just so I can formulate --

14 Q Yeah. Just if you could describe the policy and its application generally for

15 the record -- the purpose of it, what types of content it covered, and how that content

16 might've been actioned.

17 A Yes.

18 Q We can scroll down, if that would help as well.

19 A Yes. Actually, if you could scroll up just a little bit, because that would

20 include -- yeah -- the applicable policies that are included within this document.

21 So this would have related to protests that were happening after the election, as

22 well as any calls for interference within the election that related to violent extremist

23 groups, civic integrity, and, again, those violence policies -- sort of, violent threats,

24 sources of harm, glorification of violence, and coded incitement to violence as well.

25 Q And so, I'm interested in your experience with this policy guidance, both its
 19

1 development and its implementation, since you were on a team that was very close to

2 those issues. And so, if anything about it is unclear to you, just feel free to say so,

3 because your understanding of this document is an important part of, I think, the way it

4 would have been implemented.

5 So I'm going to ask some specific questions about it, and I'd just appreciate it if

6 you could try, to the best of your recollection, to talk through your understanding of what

7 the answers are.

8 So the first one is: Did the integration of the coded incitement policy into this

9 document change the application of the coded incitement policy in any way that you

10 could remember that would be relevant?

11 A Did it change the application of coded incitement in any way?

12 Q And I believe the coded incitement language is toward the bottom of the

13 document, so if we might scroll down a little bit.

14 A Scroll up a little bit. Okay.

15 Q I think page 8 of the PDF has relevance.

16 A Yes. Yes.

17 Okay. I was not as involved with the formulation of this specific document, so I

18 cannot speak to its exact creation. I can speak to my understanding after having seen it

19 and worked with it.

20 Again, I was unclear as to how the coded incitement to violence policy was being

21 wrapped into this post-elections guidance. As you I believe can see, the language itself

22 changes. I don't believe that "coded incitement" -- the language "coded incitement"

23 even appears within here.

24 Do you mind scrolling down a little bit?

25 So, if I remember correctly, the coded incitement language was essentially folded
 20

1 into -- if you scroll up a little bit more -- what -- I'm sorry, scroll down. Right there. "If

2 the content is wishing, hoping, or calling for promoting or stating identifiable groups."

3 I believe -- I believe that the information and the coded incitement language was

4 essentially meant to be represented in this, and this was kind of what we -- not I -- but it

5 was what was called a "wish of harm."

6 But, again, as you can read from this language, the full-blown, nuanced coded

7 incitement to violence policy that I had helped create and work on with other team

8 members had lost its robustness by the time that it was incorporated into this document.

9 And, after reading it, I again was confused as to what to do with those coded calls to

10 violence that we were seeing on the service.

11 Q And, just for the record, we see here in the document there are examples of

12 language that might have been actionable. For example, "The time for standing back

13 and standing by is over. Let's roll." Or, "Protests are happening on South Street. Get

14 locked and loaded. You know what to do." And those would have been bounced with

15 a strike for violation of wish of harm policy.

16 But there's not -- there doesn't appear to be language on coded incitement. Is

17 that right?

18 A Yes. So the -- yes, exactly.

19 So the language "standing back, standing by," "locked and loaded," those were

20 very specific calling phrases and turns of phrase that we had been seeing used within

21 coded incitement to violence. However, as you see, again, within this new policy

22 guidance, rather than having the nuanced approach to coded incitement to violence, in

23 which it was assessed by the safety policy team and the outcome was an abuse one-off

24 with no strike, this guidance was calling for a completely different action, right? This is

25 calling for a strike for another policy.

 21

1 Q And would you say that it's a narrower interpretation? I mean, the specific

2 phrases that are used and the one sentence at the top describing the policy. Was this a

3 harder policy to enforce against what you would've seen as harmful content?

4 A What do you mean by "harder"?

5 Q Well, less flexible, less nuanced, less --

6 A Yes.

7 Q Yes. Could you say more --

8 A Yes. So this is much less nuanced than what we would've written.

9 believe the coded incitement to violence policy would've included examples of many

10 more phrases that we would have been able to look at and evaluate under than just,

11 again, "standing back" and "locked and loaded." Those were two of the most popular,

12 but there were phrases like "Civil War part 2," "1776," "revolution," all of those turns of

13 phrase, that were also included within the coded incitement to violence policy that were

14 not incorporated underneath this post-election guidance.

15 Q And these were -- so, just as a reminder, this document applied to the

16 immediate post-election period. Was this change -- were either of these policies

17 announced publicly?

18 A No.

19 Q How did that influence Twitter's decisionmaking, the lack of a public policy

20 against incitement to violence?

21 A Yes. It was the recommendation coming from the safety policy team,

22 again, that if we were going to apply or enforce a policy that was not public information,

23 that we should not penalize users for that. Yes.

24 Q What prevented Twitter from publicly issuing a policy of this nature?

25 A Policy updates are highly complicated and take a tremendous amount of

 22

1 time to go from creation through incubation to implementation around the world. And

2 so I believe there was a time constraint.

3 And there was also the reality that Twitter's -- all of Twitter's policies around

4 incitement were in flux at this specific moment. And so we were working on what

5 would have been a public update to the incitement policies that eventually did launch at

6 the end of January, or middle/end of January. And so that would've been the public

7 update that we would have been working towards.

8 Q How long -- so this policy was in force in this fashion, according to this

9 document, all the way through the post-election period and past 1/6? Was this the

10 document that would've been in place on January 6th?

11 A Yes, this would have been the document that would, I believe -- this is one of

12 the documents that would have been in place leading up to January 6th.

13 And, again, I will reiterate, once this document was created, there continued to be

14 questions from my team specifically around coded incitement to violence, because this

15 document did not give us clear answers on what to do with that content.

16 Q What kind of questions?

17 A What we should do with the content and how we should enforce it, or if we

18 should enforce it, if the content should remain on the platform.

19 Q So there were pieces of content, presumably, that you felt wouldn't have

20 been actionable under the wish of harm policy but would have under coded incitement

21 and fell in that sort of gray zone?

22 A Absolutely.

23 Q And we'll ask more questions about this later, but, for the record now, I

24 assume you sought guidance from your manager and from Twitter leadership on those

25 important questions?
 23

1 A I absolutely did. I asked at several different times and occasions for clarity

2 specifically around coded incitement to violence and how we should be approaching that

3 content on the service, especially as we saw that content continuing to grow and rise.

4 Q And in the 2 months between election day and January 6th when the United

5 States Capitol Building was breached, did you receive that guidance?

6 A Which guidance?

7 Q Guidance on the implementation of this policy.

8 A No. No. I remained unclear and concerned from that time up until the

9 day of January 6th that, because we did not have clarity and we did not have clear

10 policies and enforcement procedures in place, that, come January 6th or come a time in

11 the future, that we would see the implication of these individuals having said, again, that

12 they were ready to -- they were locked and loaded and ready for violence. I believed

13 that violence would happen.

14 Q Was the possibility of violence brought up during any of the other marches

15 or protests between the election and January 6th -- for example, protests at State

16 capitols, the Million MAGA March? Around those events, did these conversations about

17 coded incitement reoccur?

18 A Can you remind me again, what's the timeframe that you're talking about?

19 Q These would've been in November and December.

20 A Yes. And, again, can you repeat your question?

21 Q There were multiple instances of protests which I would say had a large risk

22 of violence around which one might expect harmful statements to circulate on line. And

23 I'm wondering if, each time there was an event like that, which could be sensitive, you

24 had recurring conversations about the lack of clarity.

25 A I believe so. Yes. Again, I don't remember the exact protest or the exact
 24

1 instances, but there were continued conversations. Again, as we saw the situation on

2 the ground intensifying and I believe as we saw these protests growing, and not just

3 growing but individuals becoming increasingly armed and the rhetoric becoming

4 increasingly more violent, I became increasingly more concerned. Yes.

5 Q And during this period of growing concern, lack of clarity, what could Twitter

6 management have done to help? Like, what answers could they have given that would

7 have better equipped you really to do your job and to write and implement good policy

8 around this problem? What would've helped?

9 A Approving the policies that we had written and put in place.

10 Q So it was simply the lack of clear approval for the preexisting, more nuanced

11 policy that kind of kept you in a state of uncertainty. Is that right?

12 A Yes. Absolutely.

13 Q Okay.

14 And then this is I think my last set of questions before I see if _ , a s any

15 followup, but how liberally did Twitter apply the post-election protest guidance? Was

16 this sort of a spirit-of-the-policy application, or did they try to keep it pretty within the

17 boundaries laid out here, which are, as you described them, more narrow?

18 A My recollection is that my team, the safety policy team -- so this specific

19 section of this policy -- I do not remember a lot of enforcement happening underneath

20 this policy.

21 And what I remember, again, was more questions, multiple questions, again,

22 about content that might not necessarily have fit squarely within this post-election

23 guidance but would've fit underneath the coded incitement violence, that we were

24 wondering what we could do or what we should do with that sort of content, again, with

25 the fear and the concern that it would lead to violence.

 25

1 Q This might be a good time, if you could give us just a high-level overview of

2 how content moderation on Twitter works.

3 I understand there's -- I think Twitter Services is where frontline content

4 moderators sit in the organization, and then you are somewhat higher up in the process,

5 giving them guidance. Is that right?

6 A Yes. So there's actually a couple of different layers of content moderation

7 within Twitter. There's algorithmic content moderation, and there's also human

8 content moderation. Those often work in practice with each other.

9 The very, very first level of content moderation is mostly algorithmic, which is a

10 script -- I was not involved with writing those algorithms or understanding those -- a script

11 that searches for a lot of the most egregious content.

12 Reports -- so content that is on the service that is reported by individuals -- goes

13 to content moderators around the world. And so these are individuals who are hired by

14 third-party companies and often sit in countries all around the world who sit in what is

15 called "queues" and spend their days -- they have a certain amount chime that they are

16 given, and responsible for making decisions on whether content is a violation or not a

17 violation.

18 There's individuals, if they have questions or if there are gray-area pieces, they can

19 continue to escalate content. Once they escalate content, it then comes into, most

20 likely, a Twitter full-time employee, or an FTE.

21 The FTEs at Twitter include what you mentioned is the team called Twitter Service,

22 and so those are the safety operations team of content moderation. And so those are

23 the individuals who also tend to sit within queues and answer those escalated questions

24 and, again, make the calls and determinations, when they can, of whether something is a

25 violation or not a violation.

 26

1 If those individuals within Twitter Service cannot make a decision or if there is

2 something that is too far of a gray case or if it was a high-profile individual, that would

3 then come to the safety policy team if it was, again, within our policy scope and within

4 our policy areas.

5 Again, the safety policy team, we wrote the rules that were external. We also

6 wrote the rules that all of these other individuals used in order to make these

7 determinations. And, again, if the content was gray and no other decision maker could

8 decide -- we were not decision makers -- if no other recommenders could come to a good

9 recommendation, it would land on our desk. And, again, in very high-profile cases, we

10 were also asked to assess any other recommendations that had come in and lend our

11 recommendation as well.

12 So the safety policy team was the highest-level content moderation team and

13 system within Twitter. We were not decisionmakers. We were responsible for making

14 recommendations to our decision makers in those instances.

15 Q And who were the decisionmakers?

16 A Yes. Depending on the content, there were -- if it was not super-high-level,

17 it was not going to cause a controversy, my manager sometimes could approve. If it was

18 a little bit more high-level, Del Harvey would approve.

19 If it was a renowned politician or, say, for instance, the former President of the

20 United States, that would have to go to Vijaya Gadde, who is the chief legal officer, as

21 well as Jack Dorsey, who was the former CEO.

22 Q Okay. That's very helpful. Thank you very much.

23 I'm going to propose that - a s k any followup questions he has and then

24 maybe we can take a short break.

25
 27

1 Sure. Thank you,

2 And, J. Smith, I have a few more questions on exhibit 2, and then, a s -

3 said, we can take a break.

4 BY

5 Q So I'm curious about the existence of this document, post-election protest

6 guidance. Is it common for Twitter to put out internal guidance for events like protests

7 or other major events?

8 A Yes, it is highly common.

9 A large part of the safety policy team, in addition to everything that I have

10 mentioned, is, if and when there was an event happening in the world in which our

11 policies would be implicated -- say, for instance, large -- this happened with large protests

12 that happened all over the world, whether they be in Hong Kong or France or India or

13 within the United States -- if we saw and knew that these events were taking place, we

14 would issue what was called enforcement guidance or we would issue what was called a

15 resource request or a sweeping or monitoring request.

16 And that essentially was a request that said -- it would give a background, and it

17 would say, because of these specific situations and circumstances that we know to be

18 happening on the ground, we want you all, the content moderators around the world, to

19 be aware that this is going on. And it would also give specific directions that said, in

20 these instances, this is how we want to interpret specific policies in these specific

21 examples and in these cases.

22 To be honest, oftentimes we issued guidance in this sort of area if we were

23 making a policy change or an update very quickly because we saw, again, the need

24 because of violence occurring on the ground or the potential for violence occurring on the

25 ground.
 28

1 But it was very normal for us to create enforcement guidance whenever there was

2 a situation that was happening around the world.

3 Q That is extremely helpful context, so thank you.

4 Was it typical to do this for U.S. elections?

5 A I cannot answer that, because I only worked one U.S. election at Twitter.
 29

2 [12:13 p.m.]

3 BY-

4 Q That's very fair. Asked another way, were teams at Twitter anticipating the

5 need to put in place a election aftermath guidance prior to election day?

6 A I am not sure.

7 Q Okay. When did you first hear about the fact that you were going to be

8 asked to put together or work on post-election protest guidance?

9 A Had to be sometime after the election, but I do not -- I do not specifically

10 recall.

11 Q That's fine. And then, so this document is dated, I believe, November 5th.

12 A Uh-huh.

13 Q When you initially had conversations with other teams about putting

14 together this document, what were the concerns that they voiced and that you made

15 about the need for this guidance?

16 A Yes. If I remember correctly, leading into this guidance -- again, our team

17 had been working specifically on coded incitement to violence -- our team had collected

18 hundreds of tweets that included terms and phrases that we believed to be coded

19 incitement to violence. As -- as time progressed and we went from debate rhetoric, to

20 election, to election results being announced, to the former President declaring his

21 intentions not to concede the election, I became increasingly concerned, again, about the

22 rhetoric that I was seeing on the platform.

23 And I believe prior to this post-election guidance being created, I had a

24 conversation with Del Harvey, again, the former vice president, in which I highly and

25 strongly recommended to her that we do something about the content because I

 30

1 believed that people were going to start shooting each other, is what I told her. And so

2 from that conversation, I -- I continued to be concerned. I did not feel that leadership

3 shared that sort of same concern. And, again, this was the kind of document that was

4 created in order to realize some of those concerns.

5 Q Thank you.

6 One specific followup on that. When you told Ms. Harvey that you were

7 concerned people were going to start shooting each other, what was her response?

8 A She did not seem to believe me and did not seem to, again, share the

9 concern that this language was indeed incitement to violence. Ms. Harvey referred

10 to -- Ms. Harvey made some logic -- made -- made some sort of explanations in -- in -- in

11 these circumstances and situations.

12 And I believe the example that she gave back to me was what if someone is locked

13 and loaded inside of their house for self-defense, and sort of these sort of alternative

14 situations that were not what the clear language of the tweets was saying. And so I

15 found her response to be overly generous and reading into -- reading things into content

16 that did not exist.

17 Q Thank you.

18 Did you express concerns about President Trump's refusal to concede the results

19 of the election in light of his prior appeals to the public, as we talked about earlier in the

20 deposition?

21 A Will you repeat the question? Did I do what?

22 Q Sure, sure. So given President Trump's appeals to the public on potential

23 incitement, did you raise concerns about the potential for violence when he refused to

24 concede the election?

25 A Yes, I did.
 31

1 Q And did others raise those concerns as well?

2 A Yes. There were others on my team that raised those concerns as well.

3 Q And was this document a result of those concerns being voiced?

4 A The post-elections document?

5 Q Yes.

6 A I believe that it was the attempted response from members of Twitter's

7 leadership to address some of those concerns. And, again, it was made very clear from

8 those of us who did this work on a daily basis that our concerns were not addressed by

9 this guidance.

10 Q Okay. But to put a fine point on this, the initial draft of this document was

11 the first attempt by Twitter to respond to the possibility that President Trump's refusal to

12 concede could lead to violence?

13 A I don't know if that is explicitly happening here, but I do know that was a

14 concern of mine.

15 Q Okay. That makes sense.

16 That's all I have on this exhibit. And thank you, this has been really helpful.

17 I believe that had a followup before we go to break.

18 So, if you want to take it away.

19 Thank you,

20 BY

21 Q Hello, J. Smith. I did have one followup question I just wanted to clarify.

22 So for "locked and loaded," which you flagged to Del Harvey as a phrase that you

23 thought could in certain circumstances be a coded incitement to violence and that could,

24 therefore, be actioned by Twitter under certain circumstances, she responded to you by

25 saying, "locked and loaded" is a phrase that can be also used by someone, for example,
 32

1 saying they want to defend one's home.

2 It is my understanding, and this is the question, that often policy documents that

3 are created by your team take into account context, and that that is an important factor

4 when determining whether a violation has occurred, whether to elevate to something,

5 what sanction to apply. So I just wanted to confirm that with you and just make sure

6 that we understood that that is what you were saying.

7 A Yes, absolutely. So taking context within -- as part of the decision making or

8 the part of the recommendation process was absolutely essential for our team. So if

9 you saw, again, the coded incitement to violence policy rested a lot on what the context

10 in which a certain phrase or term was being used. So it's very, very often that there

11 was -- that was how we were supposed to approach our jobs, right, was to see what was

12 being said but to also see the context that was happening with inside of it.

13 I very much was a proponent for making sure that we used context within our

14 decisionmaking skills, especially when thinking about things like counterspeech, right,

15 which is -- very often includes a lot of language but is being reused in order to add

16 commentary or add some sort of opinion to it.

17 And so I -- I was hyper aware of those types of situations. And so if it was a

18 situation like Ms. Del -- like Ms. Harvey was saying in which a term or phrase was being

19 used in a way in which it was not harmful, or which it was not violative, or in a way which

20 it was not going to lead to any issue, I very much would've been a proponent for leaving

21 that on the service.

22 I was surprised, again, by that defense that Ms. Harvey created, because that was

23 not the language that we saw being used on the service. That was -- the language on

24 the service was not about self-defense. It was very, very much about action and a

25 willingness to participate in violence and mass violence.

 33

1 Q And just to clarify, if, for example, Ms. Harvey had been concerned that your

2 policy as written did not permit someone to say locked and loaded in terms of defending

3 their house, one option she could've had was to have been -- to tell that to you and say,

4 maybe you should tweak the policy or amend the policy or write this specifically into the

5 policy.

6 A Absolutely. So that is -- that was very, very common, right? So within

7 these enforcement guidances that we would write, we would also include exceptions,

8 right? So we would always include what we called exceptions for counterspeech or

9 exceptions in those. And we would include -- include examples that said, if you see this

10 content but it is being used in this way, again, which adds some sort of commentary, adds

11 some sort of opinion, adds something else, Twitter elevates it in some sort of way or

12 includes some sort of political discourse, leave that content on the platform.

13 So very, very much that was -- that was how we operated, was here is the way

14 that this sort of content should be treated. If you see it in this sort of context in which

15 we know it to be okay, make sure that it stays on the platform. So very much, yes, that

16 could -- that -- that was the approach that I was used to taking.

17 Q Okay. Thank you.

18 And I believe you mentioned that some of these phrases were some things that

19 concerned you that you saw -- and I'm just trying to nail down a little bit more about the

20 timeline -- I'm sorry. I'm speaking very fast. I often get admonished for that.

21 I think -- I just want to nail down a little bit more about the timeline and about the

22 approximate volume either of things that you saw when you were developing this policy

23 or that you saw when the policy you believe could've been in place but wasn't. So I

24 just -- a little bit more specifics when you said you -- you saw some things and there was

25 enough to continue, just when and how much and how did that change.
 34

1 A Yes. So a colleague of mine was working on the coded incitement -- coded

2 incitement to violence policy, had actually worked with a member of Twitter Services in

3 order to create a bot that would surface some of this content that we were seeing on the

4 service. So we gave them, again, some of these phrases and terms in order to go

5 essentially -- the bot would essentially go search Twitter.com and collect tweets that

6 included these phrases and inject them into a spreadsheet.

7 Our team received that spreadsheet. This was, I want to say, early November,

8 and our team around the world made assessments and looked at them. There was, I

9 want to say, over 500 tweets. It was an astronomical amount of tweets. Our team

10 normally reviewed maybe five tweets a day, but our team was reviewing the spreadsheet

11 of over 500 tweets, again, because of the nuance that was required within this specific

12 policy.

13 We reviewed those tweets. We left -- it was a spreadsheet that we left in, you

14 know, our recommendations for violation or no violation as well as the reasoning for why

15 we believed those things. This -- this spreadsheet, again, of hundreds and hundreds of

16 tweets was collected, and that was part of the conversation that I had with Del Harvey in

17 saying, I believe we need to remove these tweets because if we don't, people are going to

18 start shooting each other and other folks. And it was, again, Del Harvey's interpretation

19 that we did not need to remove those tweets, and instead we needed to create this

20 post-elections guidance enforcement and that was the direction that we should go in.

21 Q And I just want to clarify one thing. You said, I believe, your friend or your

22 colleague wrote a bot to scrape some things, and that was the rough material that you

23 used to start figuring out a policy that could determine the wheat from the chaff.

24 A Yes.

25 Q And this algorithmic moderation would not have been the right approach for
 35

1 this kind of a nuanced thing, this level of development in algorithms.

2 A No. No, no, no. The bot --

3 Q Okay.

4 A The bot was highly unsophisticated, right?

5 Q Yeah.

6 A What it looked for, again, was just a phrase being used but not how it was

7 being used in context. And it was our team's -- it was our team's role then to look at the

8 tweets that contained these phrases within the context to determine if we believed them

9 to be violative of the coded incitement to violence policy.

10 Q And so you, then, as humans reviewed this bot-created list that everyone

11 knew was going to be overinclusive to help determine rules for -- to help determine rules

12 that you would recommend for this policy?

13 A I believe the policy might have already been written, but we were using the

14 policy that was written in order to determine what we should do with the tweets that we

15 had collected, again, which were overbroad and overreaching. There were some that

16 were not violations within there, and we did write that down as well, but the vast

17 majority of them were violative content.

18 Q Okay. I think -- I don't think I have any other questions on this, so I'll hand

19 Thank you.

20

21 And thank you, J. Smith. You've been very helpful so far. We've been at this

22 for almost 90 minutes, so what I'd propose is that it's -- we take maybe a 20-minute break

23 and reconvene at 12:50. It's lunchtime here on the East Coast, and you may want to

24 refresh your coffee or grab breakfast. Does that sound okay to everybody, 12:50?

25 Ms. Ronickher. Uh-huh.

 36

1 Okay. Can we close the record? Jacob, do you want to do that?

2 - Sure. I'm not sure how to do it.

3 We can go off the record at 12:31.

4 [Recess.]
 37

2 [12:51 p.m.]

3 - We can go back on the record at 12:51 p.m. And at this point, I'm

4 going to hand it over to

5 Okay. So thank you, -

6 BY
7 Q We're back on the record, and I want to pick up where we left off. I have a

8 couple of general questions, and we'll turn to another exhibit.

9 I guess my first question is, during the election, what forms of technical support

10 did your team receive from other parts of Twitter? You mentioned sort of automated

11 sweeps that polled tweets into a spreadsheet that you analyzed. Were there other

12 forms of technical support that you used?

13 A Yeah. I do not believe so. I believe the only technical support that we

14 really used, especially during -- at least that I was privy to during the election was the

15 creation of those bots to identify coded incitement to violence on the service. That is

16 not out of the ordinary. Our team was mostly human capacity and capabilities and not

17 very highly technologically supported.

18 And I did want to go back to something about coded incitement that I was

19 thinking about while we were on break, if possible.

20 Q Yeah. Yeah, please. Go right ahead.

21 A Yeah. So I was thinking a little bit about, like, on the coded incitement to

22 violence as well as the post-election interference and the relationship between those kind

23 of two, and I want to just say very, very clearly, the coded incitement to violence policy

24 was never approved. It was never officially approved. It was never officially what we

25 called shipped, which meant that it was approved and sent out to content moderators
 38

1 around the world in order for them to enforce under.

2 And instead, of course, as we said, the post-election interference guidance was

3 created. I maintained significant concern that the post-election interference guidance

4 was not sufficient and actually wasn't adequate to capture a lot of the content that we

5 saw and were identifying as coded incitement to violence, which is why I continued to ask

6 questions, especially of my manager and supervisor, for clarity about what we should be

7 doing on this content that was coded incitement to violence, which I knew to be

8 extremely harmful.

9 I remember specifically at one point, I asked these questions several times on

10 Slack, in our team channel, and at one point my manager came back and said, we do not

11 have a coded incitement to violence policy full stop. And so it was very clear that, not

12 only was this policy not approved and not only was I given specific direction from the vice

13 president not to enforce any content underneath this policy, it was also told to me by my

14 supervisor that the policy did not exist.

15 Q Could we pull up exhibit 3 really quickly? There's something on that I

16 want -- I would like to follow up on just immediately here. If we could scroll down to

17 around page 12, November 4th.

18 This document is a time line of Twitter actions related to the election and

19 January 6th, provided to us by Twitter's counsel, prepared specifically for the select

20 committee.

21 If we could scroll down to -- closer to November 4. Yeah, page 12 at the bottom.

22 I'm using the document page numbers. Sorry.

23 So November 4th, 2020, the safety policy team updates the coded incitement

24 guidance to permit action on content that does not meet all of the verification criteria if

25 Twitter observes sporadic instances of unrest. From November 4th to November 9th,
 39

1 Twitter works to surface and review tweets that might include coded language with the

2 potential to incite violence.

3 So is this paragraph as written, does that capture, in your view, what Twitter did

4 during this period? And I'm also curious as to why they've bounded this language

5 between the 4th and the 9th.

6 A This -- to my reading, this information is inaccurate. We were not

7 permitted to action any content that fell underneath the coded incitement to violence

8 policy. I was specifically given directive, not just from my manager and supervisor, but

9 also from Del Harvey, the vice president of Trust and Safety, that I was not to action

10 and/or take down any content underneath coded incitement to violence.

11 The dates between November 4th and November 9th, the only -- to my

12 recollection, that may be referencing the timeline in which our coworkers worked to

13 create the bot that, again, collected all of those 500-plus tweets that contained the coded

14 incitement to violence. But, again, I went through that spreadsheet that contained all of

15 those tweets with Del Harvey, the vice president, and was told specifically that I was not

16 allowed to action the content.

17 Q So when -- when we read here, from November 4th to November 9th,

18 Twitter works to surface and review tweets --

19 A Yes.

20 Q -- we should not read that to mean that they then removed the tweets,

21 labeled the tweets, took any action against the tweets?

22 A Yes. The tweets remained on the platform until January 6th, and might still

23 remain on the platform now, to be honest with you. I'm not 100 percent sure what they

24 were taking down. But during that November 9th to -- November 4th to November 9th

25 period, the surfacing and review happened by the safety policy team within the Google
 40

1 spreadsheet, again, in which we were reviewing the language of the tweets, saying what

2 the coded incitement language we thought that it included was, whether it was a

3 violation or not a violation and why, but we were never allowed to take action according

4 to our recommendations within that spreadsheet.

5 -And, could I ask one question?

6 Please, go right ahead.

7 -Great.

8 So just to build on that, J. Smith, the second-to-last bullet or entry on the page,

9 page 12, where -- if we could scroll up a little bit, Ms. Connelly.

10 So you see on the November 3, 2020, the final action there, it says that the safety

11 policy team develops guidance for actioning language that is -- built and basically

12 describing the coded incitement to violence.

13 So as a member of someone who was involved in this policy, is it your

14 understanding that this entry describes a policy that was not implemented at the time?

15 The Witness. Yes. To my understanding, this entry on November 3, 2020, is

16 describing the development and creation of the document that you all saw, the coded

17 incitement guidance, that was done by -- it was led by another member of the team.

18 also helped with the creation of that guidance.

19 And, again, it was never approved, it was never shipped, and we were never

20 allowed to action underneath it, especially not on November 3rd, especially not on

21 November 4th.

22 - So Twitter provided this document to the select committee as a way

23 to summarize the actions it took in response to potential coordinated calls for violence

24 around the election period leading up to the 6th.

25 That entry on November 3rd, do you consider that to be a concrete action that
 41

1 Twitter took to prevent violence on January 6th?

2 The Witness. I believe that it was an attempt that was made by several

3 individuals who are very concerned within the safety policy team in order to potentially,

4 in the hopes of staving off violence that we knew could occur. But, again, it was not

5 approved by Twitter leadership or management, and so it was not -- it was not an action

6 taken by them.

7 - T h a n k you. That's really helpful.

8 I don't know if you wanted to go to another part of the questioning.

9 Yeah. Thankyou,_

10 BY

11 Q We jumped out of order a little bit here, but if we could, I'd like to pull up

12 exhibit 5. We'll return to this document later.

13 So exhibit 5 is a chart or a series of charts, really, attached to a spreadsheet,

14 labeled, "coded incitement to violence Smyte assessment summary," S-m-y-t-e, Smyte.

15 I believe, if we scroll down a little bit, I'm -- and I realize -- if we could scroll down

16 one more page.

17 This font, while still brutally small, is less so perhaps than the previous page.

18 What I really want to confirm, is this the spreadsheet that your team reviewed in order to

19 inform the coded incitement to violence policy?

20 A Yes. So this is the spreadsheet that I was referring to that our team worked

21 with Twitter Services to use. Smyte is a service that specifically creates bots. It was a

22 third-party service that was bought by Twitter and still works within Twitter. So Smyte

23 is a company that allows you to create specific bots, is a tool that allows you to create

24 specific bots.

25 The Smyte logic that was written for these specific bots was, again, looking for
 42

1 very specific language. That language was then collected within this spreadsheet. As I

2 mentioned and as you can see, we worked to create, as you see, an assessment of

3 violation, no violation, a comment that included why we believed it to be that way. It

4 includes the text of the tweet, the initials of the individuals who are reviewing it, as well

5 as we were -- we were also collecting trends that we saw that were happening within the

6 coded incitement language. So you see kind of abbreviations here, "stand back and

7 stand by," "civil war," "locked and loaded," all of those were phrases that we saw to be

8 trends.

9 So to answer your question, yes, I can confirm that this is the spreadsheet that

10 was created in early November that our team reviewed. And, again, the conversation

11 that I had with Del Harvey was hoping to take all of these tweets that we found to be an

12 assessment of violation and removing them from the service in November in the hopes

13 that this content that was, as you can see, coded incitement to violence, would not lead

14 to violence that I was afraid would occur.

15 Q Thanks. That's -- that's really valuable context for understanding this

16 document.

17 And so several of these tweets are labeled "locked and loaded." They contain

18 the phrase "locked and loaded." And I understand that was a specific phrase which you

19 argued to Del Harvey should have been actionable under the coded incitement to

20 violence policy and which she disagreed with. Is that right?

21 A Yes.

22 Q Yeah. For the record, I'd like to read some of the examples of these tweets,

23 and then I'd like to hear your opinion on whether or not the majority of these constitute

24 self-defense claims.

25 So one of them reads: I'm ready. We're ready. Locked and loaded. This is
 43

1 America and I'm an American. We will not lay down. We will not kneel. We will fight

2 and we will vote. Let's go, MAGA country. The country is literally at your fingertips.

3 Others use other phrases like "civil war."

4 One references specifically: Because the Proud Boys are standing back and

5 standing by, who's ready for the civil war.

6 Many of them tag accounts which appear to be affiliated with the former

7 President or supportive of the former President.

8 Would you characterize the bulk of these tweets as self-defense related, or did

9 you -- did you, in your view, characterize the bulk of them as self-defense related in your

10 conversation with Del Harvey?

11 A No, absolutely not. It was my -- again, I did not review all of the tweets.

12 In this document, it was -- it was literally the entire safety policy team around the world

13 that was reviewing the tweets. But I did review it, I don't know how many, but maybe

14 hundreds of them. And from my experience and from my interpretation and having

15 done this for a very long time, at this point almost 2 years, that was not what I was seeing

16 on the service and on the platform. I was not seeing language that spoke to

17 self-defense. Again, I did see language that was not a violation and I believed

18 includes -- included some sort of commentary, but the majority of that commentary was

19 not self-defense.

20 Q And to be clear, this spreadsheet does label many of these tweets

21 nonviolative?

22 A Yes.

23 Q There are some which are labeled unsure. It appears to be evidence of a

24 data-driven assessment. Would you say that that's fair?

25 A Absolutely. This was absolutely a data-driven assessment of coded

 44

1 incitement on Twitter during this time.

2 Q And when you brought this to Del Harvey, did they point to any

3 countervailing data to reject the policy recommendation you were making?

4 A No. Again, Del's example of being locked and loaded and inside of the

5 house for self-defense, I had not seen that with any of the hundreds of tweets that I have

6 reviewed. I honestly have no idea where she came up with that information or where

7 she came up with that content or where she came up with that idea that that was even

8 on the service. But, again, it was her prevailing concern in such a way that would not

9 allow us to action these tweets that were very clearly problematic and violative.

10 Q I want to back up a little bit, and if you could reflect on the -- put yourself in

11 the mind frame of it's the summer of 2020, there have been widespread protests

12 following the murder of George Floyd, and also, on the other side of that summer, there

13 was the shooting in Kenosha, Wisconsin, by Kyle Rittenhouse. And I'm wondering if

14 those events, that tumultuous summer, informed the debate you were having within

15 Twitter about the risk of coded incitement, not -- and those events, the wider civic

16 discourse, but also specific statements by the former President about those events?

17 A They very much informed the way that I went about doing my work and I

18 very much believe the way that the safety policy team went about doing the work. As I

19 mentioned, this was not the first protest situation that we had encountered. As you

20 mentioned, during the summer of 2020, there were protests around the entire world

21 related to the murder of George Floyd.

22 In those instances, I was a part of a team that created what we called the Black

23 Lives Matter playback, and that was a document that was -- it was extended enforcement

24 guidance specifically around how to handle situations and content as it related to protests

25 during that timeframe.

 45

1 So, for instance, how would we apply our wish of harm policy in situations in

2 which individuals might be calling for the destruction of a CVS property. That is how

3 granular to the level of detail that we created within that playback to determine if there

4 were calls for violence towards property, whether there were calls towards violence to

5 other individuals, whether there were calls for violence towards police officers.

6 All of those various circumstances and scenarios were very much played out and

7 analyzed and our recommendations given. And, again, that -- that playback and

8 assessment was shipped, it was approved, and it was sent to the entire world and to the

9 entire globe.

10 And so going -- going into the events of January 6th, again, this was not my first

11 time having to monitor and/or see language that is leading up to an event or within some

12 sort of tense sociopolitical context. And the way that this other work informed me,

13 especially around kind of the protests that we saw in Kenosha, was -- I believe that was

14 very much a turning point, right? We had seen protests occurring, again, around the

15 world up and to that point, but there had not been murder or exact killing in that manner.

16 And I believe that once we recognized, at least once I recognized that that was a

17 distinct possibility and not just a possibility but that it had occurred before, and I saw

18 individuals continuing to use language that only exaggerated a violent situation, it very

19 much led me to be in a state and in a place of increasing concern that what we saw

20 happen in Kenosha would be replicated on a much grander scale.

21 Q And so to summarize, it sounds like Twitter had previously created a really

22 nuanced playback for handling situations of civil unrest and potential violence and had an

23 opportunity to learn lessons from that experience, and then declined to do so after the

24 election. Is that accurate?

25 A Yes. And I will also say, again, the reason why that playback was put
 46

1 together was because there are individuals who work on these teams who deeply care,

2 and so there were individuals who saw a need. We got together and we worked on it.

3 These are -- these are what are called individual contributors at Twitter, so not managers,

4 not members of leadership. We work together in assessing what we saw to be a

5 potential problem and creating a solution to it.

6 In the instances in the summer of 2020, that guidance was approved by leadership

7 and it was approved by our management in order for us to action under it. So, yes,

8 there was absolutely a learning opportunity in which we should have and could have gone

9 into January 6th with coded incitement to violence policy created in place and ready to

10 go.

11 Q And maybe not just a learning opportunity but precedent. It sounds like

12 there was precedent.

13 A There was -- there was precedent. And I was aghast that we were not

14 following it in any way, shape, or form, again, given what we knew to be the stakes.

15 Individuals had already died. We knew that there was a likelihood that it could happen

16 again.

17 Q And so leadership approved this policy and it shipped in the summer of

18 2020. I assume that Del Harvey was part of that review process.

19 A Yeah, Del Harvey was very much a part of the review process. She

20 approved -- yes, she approved -- she approved that playback herself.

21 Q Do you have a theory of why her stance would have been different in

22 this -- in the context of the election?

23 A I do have a theory, but it is, again, my -- it is my -- I will share my -- my theory

24 based upon my experience having worked with Del in a variety of situations and

25 circumstances that all tend to appear to be similar but had different and disparate
 47

1 decisions being made.

2 I believe that there was a hesitancy on Del's part to act when it came to

3 January 6th, given the political implications of those decisions. As you mentioned, the

4 majority of individuals or a lot of the individuals and a lot of the content contained within

5 this spreadsheet is showing political affiliation and is very much aligning with a ruling

6 political party. That was not the case in these other circumstances and situations, right?

7 And I very much believe that Del was reluctant to remove content about the

8 former President and his followers because I believe not just Del but Twitter as a whole

9 very much feared the accusation of political bias, especially from the GOP, and in an

10 effort to circumvent the accusations of political bias, they either did not make decisions,

11 avoided making decisions, or decided not to make decisions, or made decisions that

12 would in no way, shape, or form show any sort of what they thought to be political bias

13 but in reality actually had the opposite effect, right, because by not -- by attempting to

14 not show political bias, they very much took this course of action of not action that

15 allowed political violence to occur.

16 Q Inaction is also a decision, right?

17 A It is.

18 Q Thinking back to the summer of 2020, of course, the former President was

19 also very active on Twitter during that period, and there were a few noteworthy tweets

20 which caused controversy and were sometimes singled out for potential response from

21 social media platforms.

22 Do you remember any conversations about the risk of Donald Trump inciting

23 violence --

24 A Yes.

25 Q -- around the Black Lives Matter protests?

 48

1 A Oh, excuse me. Sorry. Let me take back my yes. You said risk of -- can

2 you repeat the last question?

3 Q During the summer of 2020, were there conversations about the risk of

4 Donald Trump using social media to incite violence?

5 A During the summer of 2020 --

6 Q I'm thinking about tweets, for example, related to when the looting starts,

7 the shooting starts.

8 A Oh, thank you. Yes. Yes. Yes. Okay. Thank you for that.

9 I -- I will just say, literally, for the record, I have reviewed many, many, many

10 tweets, and so giving the example is very, very helpful.

11 Yes. So that -- that tweet that included the language when the looting starts the

12 shooting starts, I believe was the very first time that Twitter made the decision to use the

13 public interest interstitial on Donald Trump's account, specifically for violating hateful

14 conduct policies, specifically for inciting fear towards a protected category. So in this

15 case, it was very much about inciting fear and potentially a fear that could lead to

16 violence towards the Black community.

17 Q And if I recall correctly, and based on what Twitter has told the select

18 committee, this tweet received a PII --

19 A Yes.

20 Q -- and engagements with the tweet were limited. Users were unable to

21 like, reply, or retweet the tweet, but they could still retweet with a comment. Is that

22 correct?

23 A I do not know about retweeting with a comment. I do not -- I -- I do not

24 remember the full mechanisms and technological availabilities of the public interest

25 interstitial, but I do know that the majority of engagements were limited with the use of
 49

1 the public interest interstitial.

2 I will also just note, the public interstitial, again, is unique in that way. So other

3 labels and interstitials, say, for instance, for civic integrity or for misinformation, did not

4 include that same technological capacity in order to limit engagement with the

5 President -- the former President's tweets.

6 Q And recognizing that this might be outside the immediate scope of your job

7 at Twitter, in your view, was that effective at limiting distribution of the tweet?

8 A There has been research that has been published that very much has said,

9 specifically around labels on Twitter, that they did not stop engagement, and very much

10 oftentimes they brought attention, especially to the former President's misinformation,

11 and encouraged users to interact with misinformation in a way that allowed it to spread

12 wildly.

13 Q I'd like to -- if we could scroll back up to the first page of this document.

14 Zooming out a bit from our conversation, it appears that, if you look at the chart in

15 the upper left-hand corner, Twitter stopped reviewing and classifying these coded

16 incitement to violence examples after the 5th, it looks like. Is that correct in your

17 recollection?

18 A Yes. We stopped reviewing and collecting these examples after the

19 meeting that I had with Del Harvey arguing that we should remove the violative content,

20 and she said that we could not and should not remove the violative content and instead

21 should move towards the direction of the post-election interference guidance.

22 Q And, in fact, it looks like this chart goes through November 9th. There is a

23 post-election spike in tweets collected, none of them are analyzed, and then the chart

24 ends. So this aligns with the dates Twitter provided the select committee, the 4th

25 through the 9th, but, really, analysis seems to have stopped as early as the 5th.
 50

1 A Yes.

2 Q Yeah.

3 A Yes.

4 Q And that was as a result of Del Harvey's kind of final decision on the direction

5 you'd be taking?

6 A Yes. If memory serves correct, there were -- there was a mad

7 scramble -- again, we were under the -- we were under the idea and direction that this

8 was a problem that we needed to tackle, especially from our legal counsel. And under

9 that impression, we acted urgently in order to collect and quickly assess this content.

10 And that sort of assumption and understanding came to a very quick end, as you

11 can see, around November 11th, when Del told us that there would -- there would -- we

12 would not be taking action on these things that were very concerning to us. And

13 that -- that kind of stopped the momentum that was moving us forward so quickly with

14 an entire team working to assess these tweets around the clock, for moving to that sort

15 of urgent situation to one in which we stepped back. Another team kind of came in with

16 guidance, and we just watched this content continue to surface on the platform while

17 asking questions about what we should and could do about it.

18 Q And so you've -- you've characterized your conversations with Del Harvey

19 about this topic. I'm wondering if your supervisor shared Del's opinions?

20 A My supervisor's ideas were a little more hard to understand. Again, they

21 did -- after -- I asked -- I would ask them several times within Slack, weeks and months,

22 pointing to content or pointing to issues and asking them for clarity, specifically around

23 the policy, what we should be doing, what we should not be doing. And they did not

24 seem equipped to be able to answer those questions.

25 Q We'll return to that theme later.

 51

1 If we could open up exhibit 6 really quickly. I have a couple of questions, a

2 couple classes of questions on that, and then I'll turn it over to - f o r any followup.

3 If we could look at page 8.

4 Do you recognize this document?

5 A Yes. This is a Slack message between myself and a member of the

6 leadership team of the site policy team.

7 Q Okay. Would you mind, for the record, just reading this message, starting

8 with the word "essentially."

9 A Yes. This is coming from December 29, 2020: Essentially, between the

10 locked and loaded messages we've been seeing for months and Trump turning to the

11 fringes, I am worried about January 4th through 6th when these groups are convening on

12 D.C. for the election certifications. I am not sure if this has been discussed or if we have

13 any monitoring in place. I feel like I've been waving a red flag on this real potential for

14 extreme violence for a while, and I don't know what else to do.

15 Q So I want to, again, back up a little bit, and if we could go to page 29 of this

16 document.

17 Given that context of you've now spent -- your team has now spent almost

18 2 months working on this issue, advocating for this policy change, I want to get a sense of,

19 again, the sort of technical resources that you were requesting, provided, or unable to

20 obtain.

21 And so this page should show minutes from a December 8th meeting of the safety

22 policy team, which says that that team had data for some of their policies and not others,

23 and that they had no data or understanding of the impact of the other subpolicies under

24 hateful content that is inciting fear, referencing violence, or the upcoming incitement T2

25 updates. It meant that we were unable to track the incitement part one update to
 52

1 understand its impact on our users. This data is then also not available for the

2 transparency report.

3 What types of data were you looking for? And it's -- would it be accurate to say

4 that you continued, after the cutoff of your previous analysis, to push for more data to

5 inform your policies?

6 A I'm going to ask you to back up a little bit. Do you mind scrolling so I can

7 read -- I need to see -- the subcategories weren't available to track iOS updates. It might

8 be on the page above that talks a little bit about -- okay, yes. Do you mind if I read this?

9 A little bit more scrolling up. Okay.

10 Yes. So this is a question actually about our capacity within tooling that we had

11 within our disposal as content moderators and policy decision makers within Twitter. So

12 PV2 is a reference to what is called Profile Viewer 2. That is the internal tool that

13 individuals who have access and the proper permissions, including the safety policy team,

14 uses in order to look at accounts on the back end.

15 It is also the tool that is used by our teams when we perform what is called a

16 bounce, so deleting -- we -- Twitter does not delete content, but locking an individual out

17 of an account and requiring them to delete content is done in Profile Viewer 2 from a

18 couple of clicks.

19 What this is saying is that, within Profile Viewer, there -- when you go to bounce,

20 you are given a variety of different selections to give the reasons why you are taking

21 down that content, and the reasons typically correlate with a policy violation that has

22 occurred.

23 What is being said within this agenda item is that, within Profile Viewer 2, there

24 are not actually all of the options that are available that represent all of the policies that

25 we had. And so we were not able to actually collect the information or the data that
 53

1 would have said, if we wanted to take down a certain policy for, say, for instance, HC is

2 hateful conduct. So we would take it down for hateful conduct. VT is violent threat,

3 right? That would have been a sub -- a pull-down in a subcategory that you could've

4 clicked.

5 What that would've allowed you to do is go back in to Profile Viewer 2 and you

6 could've searched to see all of the enforcements that were actioned under hateful

7 conduct/violent threats. That sort of categorization, again, did not exist for all of our

8 policies. So, for instance, incitement of fear did not have that sort of drop-down

9 category, and so there was no data collection that was happening at the point of violation

10 or actually at any point about the content or how much content or what content was

11 being taken down under incitement of fear.

12 I believe this sort of category, this sort of question within this agenda item is

13 asking for more technological capabilities. We are asking and seeing if it is possible for

14 us to have engineering or some other team work within the tool to create the ability to

15 give us the subcategories such that data could be collected within these specific areas.

16 Q Did you receive that --

17 A Yeah.

18 Q -- tooling?

19 A No.

20 Q Was a reason given?

21 A No. But this was -- this -- that is not unusual. The current -- the state of

22 operation of tooling of data collection, of understanding of Twitter policies was

23 rudimentary, at best, at this time. And it was very, very common for us to make a

24 request for some sort of tooling solution or for some sort of additional technological

25 capacity for us to be able to do our job better, but for it to not be prioritized by our
 54

1 engineering teams such that these things were never created for our team and we were

2 just making do.

 55

2 [1:32 p.m.]

3 BY

4 Q To whom would you have made this request?

5 A I believe this sort of conversation -- this specific conversation would've

6 happened within a policy team meeting, so that request would've been heard by

7 managers and leadership within the safety policy team. And if there were to be an

8 additional request, it would've gone from them to one of the cross-functional other

9 departmental teams that work with engineering and tooling.

10 Q So probably through the same set of people we've discussed in the context

11 of coded incitement?

12 A Yes.

13 Q Okay. Did not having these types of tools create blind spots for your team?

14 A Absolutely. We -- again, it was rudimentary, especially working for a

15 technology company.

16 Q Sounds like even the spreadsheet that we discussed earlier was a kind of

17 exceptional --

18 A Yes. That --

19 Q -- situation.

20 A Yes. Again, that sort of -- the creation of bots to help surface content for

21 review in order to take some sort of data-driven approach was not the typical approach

22 that we took.

23 Q So, to recap, over the summer, you had developed a playback that

24 responded to this type of problem. Twitter took the unusual step of providing you

25 engineering resources to study the problem. And then that review was prematurely cut
 56

1 short, and a policy decision was made not to pursue a response to the problem.

2 Is that an accurate summary of our conversation so far?

3 A Yes.

4 Q One other type of file that's kind of mentioned throughout several of the

5 documents in the select committee's possession are guano notes, which seem to be notes

6 related to enforcement actions on specific accounts. Is that right?

7 A Yes. So guano notes are within Profile Viewer 2, the tool that we were just

8 talking about. So guano notes refer to a section within Profile Viewer 2 where you can

9 view all of the enforcement actions that have occurred underneath a specific account.

10 So, if you were to click on the guano notes, you would see what strikes had been

11 issued and, again, for what reason. So you would see "strike one," with the sort of

12 categorization of either, you know, hateful conduct, episodic -- whatever the case,

13 whatever the policy violation that happened. Yes.

14 Q Okay.

15 A Actually, can I note something in there too?

16 Q Yeah, yeah. Please.

17 A So I will just also say, within guano notes -- you're asking about technical

18 capabilities -- those were also highly non-sophisticated.

19 So, for instance, when it came to things like strike count, depending on the policy,

20 there was a differing strike count that would -- depending on the strike, would increase

21 the levels of severity that happened to an account. And, in certain instances, after a

22 certain amount of strikes, you were eligible and you received permanent suspension,

23 right?

24 In order to make the determination of how many strikes an account had, there

25 was no automated math happening within Profile Viewer 2. Instead, an individual on

 57

1 my team or another team would have to go into Profile Viewer, scroll through guano

2 notes that could be pages upon pages upon pages, with a notebook on the side, making

3 notes of the date in which a strike happened and literally manually calculating how many

4 strikes had occurred. There would be times when strikes were taken off of a record and

5 weren't necessarily tallied.

6 All of this to say, it led to many mistakes because of the fact, again, we were

7 manually calculating and we would not always be able to see everything within guano

8 notes or we might miss when a strike had been taken out or things along those lines.

9 So, again, to say around the technical capabilities, they were elementary and

10 rudimentary.

11 Q Thank you. That's very helpful. That's really useful context for

12 understanding some of what goes on behind the scenes.

13 I, at this time, just want to turn it over to - o see if he has any followup

14 questions. And then if you want a break, just let us know, or we can press on with the

15 next exhibit.

16 - Great. Thank you,

17 And hi, J. I did have one or two followups on this exhibit, and then we can go

18 back and I'll kind of walk through the time line that Twitter had provided to the select

19 committee and ask some more questions on that.

20 So my main followup on this compilation of documents is about the

21 December 29th email -- or Slack message, rather, that we walked through earlier.

22 And if we could scroll back to that? I'm trying to find the page.

23 That should be page 6.

24

25 Oh, no, page 8. My apologies. Page 8.

 58

1 I made the same mistake as y o u , _

3 Q So you write in this Slack message that you are not sure this has been

4 discussed -- "this" being the potential for violence in D.C. between January 4th and

5 January 6th, 2021-- or if we have any monitoring in place.

6 This is less than a week before the timeframe in question. At this point, if you

7 were going to have special circumstances in place, special protocols in place, would you

8 have known?

9 A I would have known what my team was doing.

10 The individual that I was reaching out to on this team was actually one of the most

11 senior people on the team -- our sister team that dealt specifically with terrorist

12 organizations, extremist groups, and child sexual exploitation. And so, knowing that we

13 didn't have anything happening on our side, I was reaching out to this individual in hopes

14 that maybe their team, given, again, the, sort of, violent-extremism aspect of this, might

15 have some sort of monitoring in place or might have something going on.

16 Again, I wrote this message -- as you can see, I literally say, "I don't know what

17 else to do." It was, as you said, a week before I am saying I firmly believe on these dates

18 something very, very bad is going to happen. And I very much wanted to at least make

19 sure that the leaders and leadership that I knew should have been thinking about this or

20 had the decisionmaking power to be doing something about it were aware of the red

21 flags that I was raising and wanted to make sure that they were, again, raised in their

22 vision so that those individuals would hopefully do something to address this issue.

23 Q Did you ever receive indication in the following weeks that those concerns

24 were being addressed?

25 A No, I did not.

 59

1 Q All right.

2 Well, I think we can pause there, and we'll get back to the actual immediate

3 run-up to January 6th and January 6th itself soon. But, for now, can we pull back up

4 exhibit 3,

5 Great. Thank you.

6 So, as we already mentioned, this is a document that's been provided to the select

7 committee, prepared for the purposes of responding to some of our questions about the

8 actions Twitter took in the period leading up to the election and following the election.

9 So we already talked about what was on page 12 and the bottom two posts there,

10 and just wanted to see if you had any more to add on those points, but it seems like we

11 covered the general sentiment.

12 A Yes, I believe we covered everything on those two.

13 Q Great. So, in that case, we'll move over to page 14.

14 And so, yeah, you can stop right there.

15 So you see at the bottom there, the last main entry is about Twitter's

16 implementation of a new super-spreader initiative. And, reading from the document,

17 Twitter writes that this "imposes new account-level de-amplification for

18 nongovernmental accounts responsible for repeated violations of the civic integrity policy

19 based on misleading information."

20 Can you translate that into layman's speak about what the super-spreader

21 initiative was meant to do?

22 A I wish that I could. However, this was an initiative that was created by

23 the -- so it says the civic integrity policy. The civic integrity policy was owned by the site

24 integrity team, and it was not owned by my team. So I actually do not have any insight

25 into -- this is my first time hearing of a super-spreader initiative. I cannot speak to this,
 60

1 unfortunately.

2 Q Okay. No, that is good to know. So we had a couple of questions about

3 it. I might ask them just in case you happen to know, but, of course, if you don't, just

4 say so.

5 Did you ever have indication that there were efforts being made to de-amplify

6 tweets by former President Trump in this time period, either through this super-spreader

7 initiative or any other initiative?

8 A To de-amplify the former President's tweets?

9 Q Uh-huh.

10 A No. There was never an effort to de-amplify anything from the former

11 President.

12 Q Okay.

13 And what about accounts that were closely linked to the former President that

14 were also spreading disinformation related to the election results?

15 A I was not privy to that information.

16 Q That's really helpful.

17 So, in that case, we can go to page 15. And you might also have less of an ability

18 to speak to this. But the second full entry on the page, November 8th to January 6th,

19 Twitter writes that President Trump's tweets violated the civic integrity policy 102 times

20 and that Twitter applied labels to those tweets.

21 That would've fallen under the site integrity team and not your team?

22 A Yes? Yes. Yes. So, if this -- if -- yes, the violations of the civic integrity

23 policy would have fallen underneath the site integrity team.

24 However, the majority of those -- I don't know if it was the majority, but a lot of

25 those assessments were done within a Slack channel that I was a part of. And so I did
 61

1 see some of those occurring in real-time, and I was involved in some of them if they were

2 related to my area of expertise. But I was not involved within the 102 civic integrity

3 policy violations.

4 Q Got it.

5 So, given that you were involved in some of those conversations, could you walk

6 us through the conversations within Twitter about how to take action about President

7 Trump's violative tweets in this time period?

8 A Yeah. Absolutely.

9 So, not just during this time period, but for the entire time that I worked at

10 Twitter, we were constantly reviewing the former President's tweets. The way that that

11 would typically occur, we would receive what we called an escalation, so a question

12 either from a member team or from The New York Times or from some sort of outside

13 source asking specifically about a tweet from the former President and whether or not it

14 violated Twitter's rules. Again, having done this for so long, most of the tweets that he

15 tweeted we ended up just evaluating.

16 The way that that would work, again, we would either be notified that he had

17 tweeted or they would come into our inbox, either our email inbox or our Slack in box.

18 We would be asked to assess the tweet.

19 The way that an assessment would happen is that we would essentially say, this is

20 the policy that we are looking at and that we are applying in this circumstance. Based

21 on this specific circumstance and the language of this specific tweet or context that it is

22 being tweeted into, our recommendation -- we always include a recommendation -- is

23 either violation or no violation.

24 And included with the recommendation would be an enforcement action. So the

25 enforcement action would be dependent upon what the violation was, again, depending
 62

1 on the strike count. So we'd either say, you know, our recommendation is for a timeout

2 for a certain amount of hours, if they had not received all of their strikes, or for

3 permanent suspension if that was the recommendation.

4 Within that, we would also include a why, right? So we would do an analysis of

5 specifically how that context and information within that tweet was a violation of that

6 policy as we saw it, with all of the specifics.

7 Those were written up, especially during this time. During the November

8 through January time, every single one of those assessments was written. And it was

9 provided -- it was actually -- they were all collected in one single spreadsheet about all of

10 the former Vice President's enforcement actions and all of the assessments that were

11 done on his account.

12 Q And just to clarify for the record, I think you meant the former President,

13 right, not Vice President?

14 A Yes. Sorry. Yes. Former President Donald Trump, yes.

15 Q Great.

16 So can you tell me what happened after those explanations and recommendations

17 were written by your team? Where did it go, and who did it ultimately get to?

18 A Yes. So they were not always written by my team. The ones that were

19 around safety policy were written by my team. And, actually, during this time, I believe

20 that Twitter Service actually started writing up assessments about the former President's

21 tweets because they were so frequent. They also began evaluating in writing.

22 So those assessments would go to Del Harvey. And my understanding is that

23 from Del Harvey -- any decision on President Trump's account needed to be made by both

24 Vijaya Gadde as well as Jack Dorsey. And so those decisions, I believe those

25 assessments, they at least went to Del. And if there was a decision to make some sort
 63

1 of action, I believe that those were also passed on.

2 Q And was this the usual way that violations of policies by world leaders was

3 dealt with?

4 A No. No. This was -- Donald Trump -- President Donald Trump and his

5 Twitter account received exceptional treatment by Twitter. There were former and

6 current heads of state all over the world that did not require Jack Dorsey's signoff in order

7 to action.

8 I will also say literally out loud, the tool that we were talking about, Profile

9 Viewer 2, Donald Trump's account was the only account that my team as well as every

10 other team that had access -- we did not have access to his account. We were literally

11 locked out from viewing his account on the back end.

12 And, from my understanding, if anybody attempted to search for his account,

13 there was an alarm that actually went off inside of Twitter that updated and let

14 individuals within Twitter know that somebody was attempting to look at the President's

15 account. That's how extraordinarily protected and how off-hands our team was and not

16 allowed to touch his account or his profile.

17 So, to answer your question, no. Other heads of state -- the majority of those

18 decisions, I believe, were to go to Del. They might go up to Vijaya. But Donald Trump

19 was absolutely a special and unique user on Twitter.

20 Q You touched on this a little bit w i t h - b e f o r e , but I think it's

21 worthwhile to ask again here. Why do you think President Trump was given such

22 exceptional treatment by Twitter?

23 A President Donald Trump was the Twitter platform's main attraction. He

24 was the most power of power users. And he generated users, he generated traffic,

25 which generates money. And I very, very much believe that President Donald Trump
 64

1 was good for business for Twitter.

2 And in addition to just being good for business and being an outsized user of the

3 service, I believe that Twitter relished in the knowledge that they were also the favorite

4 and most used service of the former President and enjoyed having that sort of power

5 within the social media ecosystem.

6 And to rein him in and/or find his language to be in violation would very much

7 have put those things at jeopardy. I found a willingness from Twitter leadership to make

8 logical leaps when it came to explaining very clear language from Donald Trump, a

9 willingness to read words into tweets to -- not to be able to analyze context or to create

10 context, to move and shift the rules however they should and/or could and/or needed to

11 be done in order to allow him to maintain his status on the service.

12 I also believe, again, there was an overriding and overarching fear that Twitter was

13 seen as being politically biased, especially being liberal-leaning. And so there was a

14 hesitancy to feed into that conversation or to provide fuel for the fire of the accusations

15 of censorship of the GOP by taking actions on accounts that were linked with them.

16 I will give an example of this.

17 So, very, very early on during my time at Twitter --1 believe it was a couple

18 months in -- I sat in on a presentation that was done by the research and the data teams

19 that was displaying and showing where the majority of the abuse on the service was

20 coming from. And according to the data and research teams in the presentation that I

21 saw, the underlying similarities and the networks that tied these super-abusers together

22 were very often things like "#Trump2020" in the bio or "#MAGA" in the bio or

23 "#WhereWeGoOneWeGoAII" in the bio.

24 And I remember sitting in that conversation and having someone ask, "Well, how

25 will we be able to differentiate the GOP from abusers?" And there did not seem to be
 65

1 an ability to differentiate the two, because they seemed to be one and the same.

2 And I can say that, after that meeting, I never saw that research again. It was

3 never spoken about again. There was no action taken in order to attempt to remove the

4 abuse and toxicity, even though they knew exactly where it was coming from. And I

5 believe the researcher who presented it left the company within a couple months after

6 that.

7 Q So, in the first instance, when you were sending up recommendations in the

8 post-election time period about President Trump's tweets, were there times when you

9 thought that your team's recommendations were unjustly ignored or overruled by folks

10 higher up on the food chain?

11 A Yes. Absolutely. You mentioned post- -- just post-election?

12 Q If you have examples pre-election, I think that'd be really helpful, but I also

13 wanted to focus on that timeframe after the election.

14 A Yeah. So a very, very good example I have was actually the time that the

15 safety policy team first recommended that we use the public interest interstitial on the

16 former President, Donald Trump's account. I believe it might have been -- I don't

17 remember the timeline. You'll have to look this up. But it was at the time when

18 Donald Trump was tweeting about members of the "squad," so women of color within

19 Congress. And he tweeted something along the lines of telling them to go back to

20 where they came from.

21 Our team saw these tweets. We assessed them. And, under Twitter's policies

22 and specifically within examples about abuse towards immigrants, we included the

23 phrase, "go back to your country" or "go back to where you came from." And so it was

24 the recommendation that came from our team that those -- it was a series of tweets that,

25 again, specifically identified members of the squad and specifically used this horrible
 66

1 language towards them. It was our recommendation that those tweets were, in fact, in

2 violation of our hateful conduct policies and that we should apply the public interest

3 interstitial in that case.

4 It was the response that came back from Del Harvey, that I believe she said,

5 "Frankly, I am shocked at this recommendation," and all but brushed us aside and acted

6 as if our recommendation was ludicrous.

7 And, in fact, she ended up writing her own policy assessment to find the tweet not

8 in violation and came up with -- again, when I say logical leaps, I mean the analysis and

9 assessment that Del Harvey gave us was that there could be a circumstance or a situation

10 that you could read these tweets and see them as saying that Donald Trump was saying

11 that these individuals should go back to the country that they came from in order to learn

12 some lessons about politics or about being a politician that they could then bring back to

13 the United States and then implement as Congresspeople within the United States.

14 It was absurd.

15 Q Are there any examples of that that come to your mind in the post-election

16 timeframe?

17 And I think part of the reason I am curious about that is because we started our

18 conversation, sort of, in the moment at the end of September when you became

19 concerned and voiced concerns about President Trump making more direct appeals,

20 inciting appeals to the public. And so I'm curious if in that frame -- October, November,

21 December -- there were other instances like that.

22 A Off of the top of my head, I am not remembering specific ones. That does

23 not mean that they did not exist.

24 I know that January 8th was very much an example of post-election tweets by the

25 President, the former President, that Del and I debated contentiously.

 67

1 There were instances that I can remember that were not by Donald Trump but by

2 other Congress folks and Senators. Would that be helpful to speak about?

3 Q I think if that goes to the same general point you were making about

4 Twitter's reluctance to apply some of these rules in a politically sensitive situation, then

5 most certainly, yes.

6 A Yeah, absolutely. So I'll give you an example of Senator Tom Cotton

7 tweeted -- he tweeted something, and I don't remember the exact text of the tweet, but I

8 remember it being highly problematic and targeting, I believe, an identifiable group, if not

9 a protected class. And, specifically, I believe it was targeting Black Americans.

10 And so, in this instance, our team read the tweets; again, recommended that the

11 Senator's content be found in violation. And we sent that recommendation to Del

12 Harvey, along with a larger team of individuals, who included the public policy team, who

13 became, honestly, in touch with Senator Cotton's office and began speaking to him about

14 the problems that we found within his content and were trying to back-channel with my

15 team as well as senior leadership within Twitter to understand if the Senator could delete

16 information within his tweet and re-tweet it in order for it to not be a violation.

17 It became such a conversation that Twitter's legal counsel stepped in and asked

18 that we --

19 Ms. Ronickher. Wait. Don't -- hold on. You don't want to talk about what

20 Twitter's legal counsel did.

21 The Witness. Thank you. Yes --

22 Ms. Ronickher. Or, what you heard from -- I should say, what you heard from

23 Twitter's legal counsel.

24 The Witness. Thank you.

25 It became a situation in which, while our team was absolutely convinced with our
 68

1 recommendation that the tweet was in violation, the determination that actually ended

2 upcoming down from the head of legal at that point was that --

3 Ms. Ronickher. Wait. Hold on. Don't speak about what --

4 The Witness. Sorry.

5 Ms. Ronickher. -- legal determination, but you can say what was done.

6 The Witness. All right.

7 Nothing was done. Absolutely nothing was done. The tweet was allowed to

8 stay on the platform without any public interest interstitial or without any understanding

9 of it being violative.

10 And, in the next couple of days, Senator Tom Cotton doubled down on his remarks

11 in The New York Times that then led to many Black employees from The New York Times

12 participating in protests because they believed that the language that he was using was

13 endangering their lives.

14 BY-

15 Q Thank you for that example. I think it was very illuminating.

16 I wanted to go on to a couple of other entries in the timeline. And we can get

17 back to this theme that I think you very helpfully surfaced in this conversation.

18 But on the point of the 102 violations of the civic integrity polity by President

19 Trump in that 2-month timeframe, I want to ask the question: Just in your opinion, as

20 an employee of Twitter who worked on these issues, if President Trump were anyone

21 else, would it have taken until January 8, 2021, for him to be suspended?

22 A Absolutely not. If former President Donald Trump were any other user on

23 Twitter, he would have been permanently suspended a very long time ago.

24 Q Okay. Well, that is a helpful answer.

25 I think at this point we can move on to page 15.

 69

1 Ms. Ronickher. I just wanted to ask --

2 - Sure.

3 Ms. Ronickher. -- J, how are you doing? We've been going for a little over an

4 hour. You're fine for going longer?

5 The Witness. Yes.

6 I hope that we can finish up the discussion of this exhibit by 2:30 or

7 so, and then I think that would be a good time for a break.

8 Ms. Ronickher. Perfect.

9 Great. Thank you, Alexis.

10 And thank you, J.

11 The Witness. Thank you, Alexis.

12 BY
13 Q So can we scroll to page 15? Oh, we're already there. So -- okay. Scroll

14 up a little bit.

15 On this first full entry, it says, for the record, I can say, "Twitter ends its use of

16 warning interstitials on tweets commenting on the election outcome but continues to

17 apply labels to tweets about election integrity."

18 So we've talked a lot about warning interstitials since the beginning of this

19 deposition. And I'm wondering if you have thoughts about the decision or if you were

20 privy to conversations about the decision to suspend the use of these interstitials on

21 November 7th.

22 A Yes. Unfortunately, I was not privy to these conversations. Again, the

23 interstitials that they're talking about were violations of the civic integrity policy, so those

24 warning interstitials and the decision to end and/or use those would have rested within

25 the site integrity team and not the team that I sat on.
 70

1 Q Okay.

2 So, if we can go down to November 10th. I think this is the third -- yeah, here we

3 go -- the third entry on that page. You can see that on November 10th Twitter writes

4 that it implemented automated measures to detect and escalate certain phrases linked to

5 violence, like "civil war" and "locked and loaded."

6 And we talked about this before, but, given our conversations around this, how do

7 you view this characterization of Twitter's actions?

8 A I am honestly not sure what they are referring to here. Again, from the

9 information that we have looked at, the Smyte role stopped collecting information I

10 believe on November 9th.

11 And to my knowledge and to what I worked on on the team, after the collection of

12 those tweets within a spreadsheet, there was not detection or escalation occurring. If I

13 remember correctly, I believe on January 5th during a conversation we even asked if we

14 had done anything with those tweets or if Del had seen them or approved them.

15 No, this characterization does not make sense to me.

16 Q So, to your knowledge, there was not automated measures that are

17 mentioned here?

18 A To my knowledge, no.

19 Q And assuming, arguendo, that there were some of these automated

20 measures -- at this point, there had been 6 months of heightened activity by far-right

21 groups related to COVID lockdowns and Black Lives Matter counter-protests -- would

22 these measures have been an appropriate response, these automated detection

23 measures, at those prior junctions of violence?

24 A Would you repeat the question?

25 Q Sure. So there had been at least 6 months of heightened activity by

 71

1 far-right groups, as we've discussed throughout the deposition, and, at those prior

2 moments, would you have wanted to see automated measures to detect and escalate

3 phrases related to civil war or potential vigilante violence?

4 A Would I have liked to have seen something like that? Maybe. And I don't

5 know -- again, I don't know how relevant it would have been until after the debate in the

6 election when things started becoming even more askew.

7 Q Uh-huh.

8 A You mentioned -- I'm gonna -- I'm gonna stop there.

9 Q I think that -- that makes sense.

10 And, I guess, if that's all we have on this one, we can move on to December 10th.

11 Oh, go ahead.

12 A Yeah. So you mentioned COVID lock-downs and, kind of, protests that

13 were happening and had been happening.

14 Q Yes.

15 A Would you like me to speak to any of my experience in working on that

16 information?

17 Q I think that, to the extent that it fits into the broader conversation we've

18 been having about Twitter's enforcement priorities and its willingness to protect from

19 real-world harm, then of course.

20 A Absolutely.

21 So, as I mentioned, part of the work that my team did was, very early on,

22 COVID-19 misinformation. And so, when COVID hit the world, so beginning around

23 March 2020, Twitter released its very first misinformation policy, ever, directed towards

24 public health harms that could occur based on the COVID-19 pandemic.

25 Very shortly after the pandemic hit and we saw mask measures and stay-at-home
 72

1 orders being given around the world, we also saw individuals beginning to organize

2 protests in order to protest against the COVID measures that were being taken.

3 At that time, it was Twitter's official stance that it was actually taking down all of

4 the content that was calling for assembly or was calling for protest or was calling for

5 gathering that could happen against these COVID measures. And it was actually me who

6 came into the conversation, specifically, again, around counter speech, around free

7 expression, around symbolic speech, around the freedom of assembly, and made the

8 argument to Twitter leadership that we needed to stop removing individuals' calls to

9 protest and to gather, because these were fundamental rights that individuals had, and

10 they very much had the right to assembly, especially to protest against their government.

11 It was a very contentious conversation in which Twitter leadership did not listen to

12 me for several weeks. And I remember specifically saying to them, we need to make

13 sure that our restrictions are content-neutral, because we cannot stop people from

14 gathering just because we do not like what they are gathering for or what they are

15 gathering to say. As we know, that is First Amendment jurisprudence. That is what my

16 training background is in.

17 And so, understanding that, I ended up having to send around an article from

18 several ACLU organizations that very much argued the exact same thing that I did. And

19 the decision was made without informing the public, without informing anybody, and we

20 started allowing those calls to be made on the service.

21 And so I mention that because, again, working with protests and understanding

22 the situation and realities and circumstances on the ground and balancing free expression

23 and safety during this time was very, very much my skill set and my expertise. And so I

24 very much went from the very beginning of this and seeing how protest went and then

25 understanding why around November/October I saw a huge shift in the conversation,

 73

1 because I had been watching these things happen for so long.

2 Q And, again -- thank you for that answer.

3 Again, you mentioned earlier t o - t h a t you saw certain hashtags that

4 were uniting a lot of the far-right activity that tended towards violence on the platform at

5 the end of 2020.

6 So, given that, to what do you attribute this shift in tenor, the shift towards

7 violence, from earlier protests and the late November/December moment?

8 A Yeah. I really believe it was just a set of socio political crises that led to that

9 situation, right?

10 Because, again, we went from March 2020 of individuals calling for peaceful

11 protest against mask mandates being issued by their government, in which some of them

12 were armed protests, which, again, is very lawful and okay, to then seeing Kyle

13 Rittenhouse shoot individuals in Kenosha and recognizing that these protests, especially

14 as they became increased with -- as they became intertangled with protest against racial

15 discrimination within this country, I believe very much began to change the tenor of the

16 conversation from rights and liberties to literally civil war.

17 And so, seeing that sort of change that was no longer "I want to go to my capitol

18 and protest my government because I don't want to wear a mask" and instead seeing "I

19 am locked and loaded and ready for the civil war part two happening on January 6th,"

20 that is a huge shift that I saw occur. And, again, going from peaceful to actually having

21 violence occur, to continuing to call for that violence to not only occur but for it to occur

22 at a grander scale.

23 Q Thank you. That is helpful context.

24 So I do want to move a little bit closer to January 6th, but, first, the last action

25 item you can see on the page right now, December 10th, Twitter tells us that the hashtag
 74

1 "civil war" was de-amplified --

2 A Yes.

3 Q -- from trends. So can you explain to us what de-amplifying a hashtag from

4 trends means?

5 A Yes. I believe I was involved in this decision.

6 So I was the point of contact for the safety team, when it came to trending topics

7 or when it came to things that trended on the platform, in assessing whether or not we

8 should deny-list specific terms or phrases that were trending, based on their implications

9 for safety.

10 And so, when we saw in, again, December the hashtag "civil war" start to trend, I

11 believe I was asked to assess and determine whether or not we wanted to deny-list it

12 from trends, which means it would no longer show up within the trending topics or be

13 recommended to individuals as a trend that they should search and/or look through.

14 While there was counter speech that very much, you know, would say things like

15 "I can't believe we're about to have a civil war" or jokes -- you know, it's Twitter; there

16 are jokes everywhere about things -- it was clear to me that, given, again, the protests

17 and uprisings that we had seen about rights for Black Americans being coupled with

18 political protest now being classified as an actual civil war, knowing the history of the Civil

19 War within the United States of America, I became increasingly concerned that leaving

20 this sort of trending topic to be seen and engaged with throughout the service would only

21 amplify and/or radicalize individuals who were already believing and/or leaning into the

22 narrative that they wanted a civil war to happen again.

23 Q Understood.

24 Were there other trends that you remember pushing for de-amplification or

25 deny-listing that higher-up individuals denied?

 75

1 A Yes? Well, I'm sorry, I'm sorry. What -- myself pushing for trends that --

2 Q Yes, yourself or others who were involved in this process pushing

3 for -- unsuccessfully pushing for other tags to be de-amplified.

4 A Yes. No, I don't.

5 I think -- and that is -- I think the majority of that is due to -- so, normally, the way

6 that we would handle these things is, again, when we create enforcement guidance, like

7 the code of incitement to violence guidance, it would include a list of hashtags that we

8 would proactively deny-list so they could never trend.

9 And part of what we would've hoped is, in writing that guidance and having it be

10 approved and having it being shipped, is that things like "civil war" or things like "locked

11 and loaded" or things like "stand back and stand by" would have never had the

12 opportunity to even begin to trend on the service, and if they did even get the numbers

13 to begin in that place, they would've not have been able to technically fill the slot and be

14 shown as a trending topic because the back end would've disallowed that.

15 Q Okay. That makes sense.

16 So now I want to move to page 16. And we just have one more entry I want to

17 go through and then one question that kind of goes back to -- goes both forward and

18 backward. But this one for now is December 19th.

19 So, on December 19th, President Trump used Twitter to post an article about

20 post-election fraud, alleged fraud, and then said, "Big protest in D.C. on January 6th. Be

21 there, will be wild!"

22 And I'm wondering if you could talk us through how you and your team reacted to

23 that tweet.

24 A Yeah.

25 So I remember seeing this tweet. And, again, having spent at this point over a
 76

1 month waving a red flag and stating again the potential for mass violence to occur, when I

2 saw this tweet, I saw it as the date and location in which the mass-casualty event would

3 occur. And it became clear to me that all of the angst and riling-up that I had seen on

4 the service that was in the air was being directed towards, again, a specific date and a

5 location. And that, for me, became -- I was already increasingly concerned; that became

6 mind-boggling to me.

7 I think we talked about the message that I sent to one of the team members

8 specifically saying out loud, these are the dates where it seems like violence is going to

9 happen. And so this tweet, for me, was an organizing and a rally call that was saying:

10 To all of you all who are locked and loaded, to all of you all who are standing back and

11 standing by, to all of you all who are ready for civil war, make sure that you are at this big

12 protest in D.C. on January 6th.

13 I will also say, what shocked me was the responses to these tweets, right? So

14 these were -- a lot of the "locked and loaded," "stand back, stand by," those tweets, were

15 in response to Donald Trump saying things like this, right? So there would be a response

16 that said, "Big protest in D.C. on January 6th, be there, be wild," and someone would

17 respond and say, "I'm locked and loaded and ready for civil war part two," right?

18 And so those were the tweets at this moment and moving into the end of

19 December that had me wildly concerned that not only was there going to be mass

20 violence but that it had a specific timeframe and location attached to it.

21 Q Thank you for that background.

22 I do have a couple of followup questions on the response to the President's

23 December 19th tweet.

24 The first is, could you expand on the nature of these responses by run-of-the-mill,

25 low-level Twitter users? How did the tenor of those posts change before and after the
 77

1 December 19th tweet?

2 A Yeah. Again, it was -- it felt as if a mob was being organized and they were

3 gathering together their weaponry and their logic and their reasoning behind why they

4 were prepared to fight.

5 Prior to December 19th, again, it was vague, it was nonspecific, but very clear that

6 individuals were ready, willing, and able to take up arms. After this tweet on

7 December 19th, again, it became clear, not only were these individuals ready and willing,

8 but the leader of their cause was asking them to join him in this cause and in fighting for

9 this cause in D.C. on January 6th as well.

10 Q So the traffic you were seeing before December 19th, were those tweets as

11 focused on the date of January 6th as they were after the tweet?

12 A They were not. So before December 19th, I did not see clear dates. I just

13 knew something bad is going to happen. After this tweet, it became very clear to me

14 that what is happening here is being organized for the weekend of January 4th through

15 6th, specifically on January 6th, because that was when the election certification would

16 be read into record.

 78

2 [2:30 p.m.]

3 A They were not. So, before December 19th, I did not see clear dates; I just

4 knew something bad is going to happen. After this tweet, it became very clear to me

5 that what is happening here is being organized for the weekend of January 4th through

6 6th, specifically on January 6th, because that was when the election certification would

7 be read into record.

8 BY-

9 Q And when you were looking at other influencer level accounts, perhaps not

10 as large a following as the President but accounts that are closely associated with the

11 President, did you also see a shift in how those accounts were treating potential protests,

12 potential violence?

13 A Absolutely. Absolutely. I remember Madison Cawthorn

14 released -- Madison Cawthorn had been like a rising start on the GOP and on Twitter at

15 that time, and was very much leaning into the idea of the election being stolen and was

16 producing a lot of flashy content to that effect that was -- had a lot of engagement on the

17 service. And if I remember correctly, after this tweet, Cawthorn released -- released a

18 long video, I want to say, that essentially summoned folks to come to the Capitol for

19 January 6th.

20 Q And I know that you were not necessarily in charge of looking at accounts

21 associated with extremist groups or far-right groups, but to the extent that you were, did

22 you see a shift in sort of bigger name accounts in that area of the universe as well?

23 A Yes. Yes. I believe any -- I don't believe -- the majority of high-profile

24 accounts that were associated with the GOP party and Donald Trump at that point began

25 tweeting about the events and converging on D.C. on January 6th.

 79

1 Q So at least in terms of the social media atmosphere, the atmosphere on

2 Twitter, do you think that January -- the Capitol on January 6th would've become as

3 clear-cut a target if President Trump had never posted this tweet on December 19th?

4 A No. I very much believe that Donald Trump posting this tweet on

5 December 19th was essentially staking a flag in D.C. on January 6th for his supporters to

6 come and rally around their -- there was -- again, that was no definitive date, there was

7 no definitive time, there was no definitive location for people to gather and to act out and

8 to engage in the activity that they had been speaking about doing before this tweet.

9 Q And you were concerned about the potential for this gathering becoming

10 violent?

11 A Absolutely.

12 Q So, with that, I wanted to shift into what you did when you saw this tweet.

13 Did you and your team escalate, as we've talked about earlier, to others at Twitter?

14 A I do not remember what our assessment for this exact tweet was. I'm sure

15 at this point we were sent this tweet to look at and see if it violated our policies.

16 Looking at it on its face, as you can see, there is nothing within there on its face value that

17 would be a violation of any policies. And so it was, you know, deemed to not -- to not

18 be in violation of our policy. And I believe we allowed it to -- to remain on the service

19 and not -- not do anything with it.

20 Q That's -- that's very interesting. So as you say, there's nothing on the face

21 of that tweet that -- that suggests direct incitement to violence or violent activity. What

22 did you think needed to be done at that point in order to prevent these coordinated calls

23 for violence from coming to fruition?

24 A I believe that we needed to implement the coded incitement to violence

25 policy, and began removing tweets from the service that had been calling for this violence
 80

1 for months.

2 Q Would you say this -- this tweet itself on December 19th is an example of

3 why the coded incitement to violence policy was needed?

4 A Yes, because I believe that it would have allowed us to be more proactive in

5 responding to the responses of this tweet, right?

6 Q Uh-huh.

7 A Because it became a pile on, it became an event, it became a gathering, it

8 became a place to be. And if we were at a point in which we would have been able to

9 see responses to these tweets in which folks were clearly calling for violence and began to

10 remove them, I believe that we would not have seen so many other individuals feel riled

11 up to violence and feel as if they should also participate in these calls to violence.

12 The coded incitement to violence policy most likely would have not impacted this

13 tweet itself, right? It was not a public policy and most likely would never have impacted

14 President Donald Trump himself, but I believe it very much would have allowed us,

15 allowed Twitter to rein in the extensive calls for violence that began -- that had already

16 been on the platform and that became even more targeted with date and location after

17 this.

18 Again, I remember seeing tweets that literally said, like, I am locked and loaded

19 and ready to be in D.C. for a civil war on January 6th. They were very, very specific.

20 Q So just a few more questions on this. Given that the calls for action

21 became much more specific and targeted after this tweet, did you have conversations

22 with your supervisor, with Del Harvey in the timeframe between December 19th and,

23 let's say, December 29th when you sent that message on your chat?

24 A I don't remember if I had conversations with Del Harvey at that point, but I

25 do remember sending a Slack message to, I believe it was the health policy team, so that
 81

1 much larger team, and specifically posing the question to them and saying, if this were

2 any other country that we were monitoring in which the results of a fair election were

3 being contended by the ruling party, and that ruling party was openly making calls for

4 violence and specifically calling for the enactment of a civil war upon that country, would

5 we be handling this any differently? And because it was my belief, because, again,

6 having done this job all around the world, that if it was any other country, we very much

7 would have put guards -- safeguards in place to make sure and ensure that these calls for

8 violence were not as prominent and that they would not have led to what we have known

9 to have been offline violence. Yeah.

10 Q Thank you. That is again helpful.

11 So I know we have to take a break soon, we've been going for almost 2 hours.

12 just wanted to, while we have the timeline up, get to two or three more questions.

13 And so if we can go to -- scroll down to January 6th at 2:20 p.m.

14 We're going to get back to other parts of the day, but you can see the second to

15 last full action item there kind of goes to everything we've been talking about. At 2:20

16 p.m., just around when the Capitol was breached, Twitter begins to identify and review

17 coded language that is potential to incite violence.

18 Do you take this as an ad hoc implementation of the incitement to violence policy

19 or coded language policy?

20 A No. I -- 2:20 p.m. EST. Trying to think back to the time line of that day in

21 my head. And I believe at that point, the march had already concluded and folks were

22 converging on the Capitol. And in that case and in that instance, I believe that this is

23 most likely referring to the enforcement guidance that myself and two other members of

24 the safety policy team created on January 6th that essentially spelled out how we should

25 and could use the coded incitement to violence policy to begin to take down a lot of the
 82

1 content that we were seeing being created on the platform on that day as individuals

2 were actually beginning to breach the Capitol.

3 Q That makes sense. So it's in keeping with our earlier conversations on the

4 fact that coded incitement policies were not actually in place yet.

5 A Yes. No, there was nothing in place. There was nothing in place.

6 Q Okay.

7 There's one followup question I have, and we will break soon, certainly within the

8 next 10 minutes. Because after the break, just a preview, we're going to go into the

9 immediate run-up to January 6th and then January 6th itself.

10 And this was asked just in terms of the nexus to coding incitement, but I had a

11 question about sort of the tracking of potential violence that you did before and after the

12 December 19th tweet. And I'm wondering if you saw traffic on other platforms that

13 were catering to pro-Trump, far right parts of the internet, thinking of Parler, any of

14 thedonald.win community sites. Did you have visibility or did Twitter have visibility into

15 those sites? And did you also get concern from the perspective of what was happening

16 there too?

17 A Yeah, that's a great question. Twitter -- Twitter itself, the majority of

18 analysis and decisions that are made -- let me back up.

19 Twitter does not -- other platforms have what they call like an off-service policy,

20 which allows you to look at what is happening on other services or other platforms and

21 use that and apply what is happening over there to circumstances that are happening on

22 your service or platform. Twitter does not have a policy along those lines. And so

23 any -- any decisions that are being made about what is happening on Twitter have to

24 relate to what is happening on Twitter and not what is happening on any other service or

25 what is happening on any other social media service.

 83

1 So, no, I did not look at Parler, I did not look at the chans, or I did not look at

2 Redd it, or any of these various -- their sources. I did keep track of the news as it was

3 coming, and so I did hear, you know, reporting about, you know, things that were being

4 said and done within those circles that made me concerned. But as far as official

5 tracking of other services or platforms in relationship to what was happening on Twitter,

6 that was not done.

7 Q Do you think it would have been helpful to --

8 A Yes.

9 Q -- to take into account?

10 A Yes. I believe that -- I believe that off-service policies are incredibly

11 beneficial and are very much needed within the future of content moderation, given the

12 realities of what we know to be the information ecosystem in which none of these

13 services operate within a vacuum.

14 Q In the run-up to January 6th, did you ever discuss the need to look at the

15 broader ecosystem of content with your supervisor or Del Harvey or anyone else?

16 A That's a great question. I don't know if I explicitly did. Well, I

17 mean -- yeah --yes. I mean, the message that we looked at that I sent to you, one of the

18 site policy leaders was -- was specifically around an article that he had shared, right? So

19 it was outside information and outside knowledge that he was also bringing in and

20 sharing. And my response to that was, thanks for sharing this. I also share a major

21 concern that was highlighted within this based upon what I have seen on our service.

22 Q Got it. That's very helpful.

23 So at this point, I want to ask for a break i f - o r or

24 anyone else has any followup questions on these points.

25 I did have one.

 84

1 BY
2 Q I just want to make sure I think I understand what you're saying and the

3 import of it. So, obviously, Twitter cares about what happens off-site -- off of the site.

4 Violence off of the site is something that Twitter was interested in and would, according

5 to their policies, try to stop. So I guess just, I want to -- I want to make sure that we get

6 a clear understanding that -- that Twitter didn't -- that Twitter, at least for your work, had

7 a policy of focusing only on what happened on Twitter.

8 A No.

9 Q No?

10 A No. No, no. Okay. Yeah, let me clarify -- let me clarify that a little bit.

11 So when I say -- when I say kind of off-site and -- and only what happened on Twitter, I

12 very much mean that in relationship to the policies about the way that the social media

13 service evaluates its content, right? So, for instance, Twitter will not look at a tweet and

14 then also go look at that individual's Facebook page and see, like, what else are they

15 posting on Facebook and is that worse than over here and should we apply that there.

16 That is not a -- that's the kind of an example of an off-service policy. That does not exist

17 at Twitter.

18 What is important to Twitter and the sort of question I think that is important to

19 highlight here, offline harm is the biggest consideration, was the biggest consideration for

20 the safety policy team. Again, our job was to weigh safety and free expression. And

21 the safety and the types of harms that we were most concerned about were bodily harm

22 and violence happening offline and off service.

23 So the biggest concern was something being said online and on Twitter and that

24 having the implication or effect of causing death or bodily harm to individuals off of

25 Twitter inside of the real world. And that was a huge purview to what we were doing.
 85

1 Q Yeah. So, for example, if they were trying to figure out what a dangerous

2 organization was and who was inside of it, they might look at off-site information to

3 determine that. But specifically for interpreting content online, you say -- what did you

4 call it? There's no off-site policy?

5 A Yes.

6 Q Okay. So --

7 A Yeah.

8 Q -- I guess, a followup question. - a s k e d about whether you brought

9 up the need for it. What was the reasoning for not including this kind of context? As

10 you said earlier, as a First Amendment expert, context is important, and so I'm just

11 wondering what the reasoning was for not investing in this.

12 A You're saying, not investing in what?

13 Q Not investing in or developing a policy that considered off-site speech --

14 A Yeah. Yeah.

15 Q -- or activity when determining content.

16 A Yeah. I think that's very much just a -- so I know about off-service and

17 off-site policies from working at other companies and having worked it with them, and it

18 is a newer type of policy and a broader type of policy than what has typically existed

19 within content moderation spaces. So it's industry-leading to have one of those.

20 There are only maybe, I would say, a handful of companies that actually have a policy like

21 that.

22 And so why would Twitter not have developed or put in resources in order to

23 develop an industry leading policy very much because Twitter was not at the point of

24 having industry leading policies. We were very much filling gaps or, you know, working

25 to align the current policies with human rights standards, right? We were playing a lot
 86

1 of catchup. We weren't necessarily able to be on the cutting edge of creating policies

2 like off-service policies.

3 Q So that makes sense. One last followup question is, was it a policy at

4 Twitter to affirmatively not consider this, or was this just something Twitter didn't have

5 an affirmative policy on and didn't -- didn't have resources dedicated to it and didn't

6 consider it, or was there like an affirmative statement somewhere where Twitter said, we

7 are just not going to have time or we're not interested or we're not doing this?

8 A I don't know that I know the answer to that question. I do know that when

9 I was being trained to work on the team and discuss content on Twitter, it was very much

10 in my training to just use Twitter to assess and not use other services.

11 Q Okay. And, you know, you were -- you -- you were supposed to following

12 your training?

13 A Yes. I was, yes.

14 Q Yeah. So they gave you the tools and they gave the instructions that they

15 felt would be necessary for you to perform the job?

16 A Yes.

17 Q And that included focusing on onsite activity?

18 A Focusing -- yes, focusing on the content that was happening on Twitter with

19 an eye to how it might lead to violence offline.

20 Q Okay. I don't think I have anything else. Thank you very much.

21 Ms. Ronickher. All right. I think at 2 hours --

22 Yes. So I think that we can go at the -- sorry. We can go off the

23 record at 2:51, and everyone get coffee, and we can come back at 3:12, 3:15.

24 Ms. Ronickher. Is that enough time for you, J.? Do you need --

25 The Witness. I'll be okay. I'll be fine. All right. 15 after, you said?
 87

1 Yes, 15 after. We could -- an extra 3 minutes isn't going to kill

2 anyone.

3 The Witness. Perfect. See you all then.

4 Ms. Ronickher. Okay.

5 Thank you.

6 [Recess.]
 88

2 [3:17 p.m.]

3 - All right. So we can go back on the record at 3:17 p.m. And at this

4 point I'm going to hand it over to to walk us through the weeks leading up to

5 January 6th and 7th.

7 BY
8 Q I think there were actually a few more follow-up questions on the last

9 portion of questioning before we proceed to the lead-up to J6 itself. That was a very

10 rich discussion, and kind of just a few loose ends that I think and myself

11 wanted to tie up, the first of which is, we touched a little bit on escalation of Donald

12 Trump's Twitter account, the way that policy decisions around enforcing on his tweets

13 would have been escalated. This seems to be above and beyond even the very

14 important tweeter policy. Is that -- would you say that's right?

15 A Yes. So we -- we had -- what you're referring to is when there were

16 individuals who were referred to as VIPs and there were certain criteria effort that -- I

17 believe it was a couple of the criteria were over 100,000 followers, had to be verified. In

18 those instances, before we took action on an account, we would send out a notice to

19 stakeholders so other folks in the company who would be responsible for parts of any

20 fallout, so communications teams are often alerted or public policy teams are often

21 alerted, would have to be in touch with those individuals crossing communications

22 offices.

23 The VIP kind of heads-up was, again, just -- it was a -- it was a kindness, really, to

24 just give our colleagues the leeway and time to be able to prepare for kind of, again, the

25 fallout or what they would need to do with their jobs based on the actions.
 89

1 The escalation of the former President's tweets went above and beyond any other

2 user on the service. Again, those tweets coming from the former President had to be

3 approved by Jack Dorsey himself in order to make those decisions. When it came to

4 VIPs, depending on who it was, sometimes my supervisor was able to make the decision,

5 sometimes it needed to go all the way up to Del Harvey. But for the majority of VIPs, it

6 did not need to go up to Jack Dorsey. Sometimes it might need to go to Vijaya,

7 depending on the individual. But the process -- the process for reviewing and handling

8 Donald Trump's account was absolutely an exception rather than the rule.

9 Q Thank you. That's very helpful.

10 So when -- as you said, the VIP policy was like a kindness to let your colleagues

11 have time to prepare to do their jobs, to handle any fallout or anything else.

12 Was there a standard delay on how long it would take to action those tweets or

13 how long it would take to get a judgment back?

14 A Yes. So -- so there -- the amount of time that it would take it to get an

15 adjustment back or kind of assessed was always dependent upon how long it took our

16 team to do those things. However, when we came to a final recommendation and the

17 recommendation was approved, we would give our colleagues what we call a 4-hour

18 heads-up. And we would get -- literally, we'd start the clock and give them 4 hours

19 before we would make an enforcement action on the account in certain and unique

20 circumstances. So, for instance, if a VIP -- this happened once, I think it was Giuliani

21 actually who tweeted someone's phone number, their personal cell phone number. In

22 that instance, we gave -- we just sent an email, a heads-up that we had already taken

23 action.

24 So, instances in which it is public information or private information that was

25 released or, for instance, if it was nonconsensual, maybe there was something that
 90

1 needed to be taken down immediately, we would take it down immediately and give

2 notice. But if it was instances that were not immediate, we would give that 4-hour

3 heads-up and wait.

4 Q And did you ever receive approval to take action before that 4-hour time

5 window was closed, or was it always just kind of 4 hours?

6 A I believe there were a couple of times. I believe even the example that I

7 gave, if I remember correctly, I think I sent out a heads-up and somebody was like take it

8 down immediately because of what it was; again, it was someone's private phone

9 number. So there were instances like that.

10 Q Okay. But ordinarily it would take the full --

11 A Ordinarily, we would allow the time clock to run for a full -- typically, what

12 we would literally do is we would include a member of Tweeter Service on the email and

13 we would say, please action this in 4 hours. And so it would be on their team to set a

14 clock. And in 4 hours, they would send an email to the entire group saying that

15 that -- that the account has been actioned.

16 Q And this 4-hour starts after you make your assessment?

17 A It's after the assessment and recommendation have been approved.

18 Q Ah.

19 A So after our team has said we recommend violation, no violation, and that

20 recommendation has gone to the decision-maker, and the decision-maker has said, yes,

21 we're going to move forward with this, then we would move forward with sending a

22 4-hour heads-up. So it would often take much more time than 4 hours after the content

23 had hit the service for it to be removed.

24 Q Yeah. How long would you say it typically takes to have an assessment

25 approved?
 91

1 A It depended. There were times when it took this instant. There were

2 oftentimes in which -- I think the situation that I was talking about earlier with Tom

3 Cotton was one of the situations in which it was determined it actually had taken too

4 long, and so we weren't going to do anything. We weren't going to take action because

5 the damage had been done, is what we were told by leadership, because the deliberation

6 period -- whereas our team had made the recommendation of violation, the deliberation

7 that had had to happen between leadership took hours upon hours upon hours to the

8 point that it was the end of the day by the time it would've -- we would've been able to

9 enforce, and leadership made the decision to override our recommendation and not take

10 action because it -- so much time had elapsed.

11 Q Wow. So did you ever -- was this ever a popup conversation that, you

12 know, by the time 4 to 8 hours or longer has passed, a tweet has lived a lot of life, can

13 travel pretty far. Did you ever talk about that being a problem?

14 A Yes. This was brought up in the email chain, especially, that I was

15 referencing with Tom Cotton on how problematic it was for us to have to wait for so long

16 and to be kind of concept of the damage being done that while, you know, a tweet might

17 have existed for several hours, it will continue to last for the rest of Twitter exists, and so

18 the damage might not have actually been done, can still continue to be done. But

19 that -- yes, there was -- there was pushback on that line of thinking and around the

20 process and how long the process was taking to -- to implement some sort of impact.

21 Q You mentioned the incident with Tom Cotton. In our previous discussion,

22 you mentioned an incident with Madison Cawthorn. Were there any other notable

23 elected officials who -- whose tweets were escalated between the election and

24 January 6th?

25 A Yes, there were many, many other. I believe I worked on a -- a tweet by

 92

1 Matt Gaetz that also received the public interest interstitial. The White House account, I

2 believe, also received the public interest interstitial. I am sure that we reviewed several

3 tweets by Ted Cruz. I do not believe any were found to be in violation. And it was not

4 uncommon for my team to spend parts and chunks of our days reviewing the timelines of

5 prominent GOP members during that time in order to assess their language, because,

6 again, it had become increasingly towards the violative side.

7 Q Okay. Given, you know, the totality of our conversation today, I did want

8 to take -- give you a chance, if you want to, to step back and answer a sort of high-level

9 question about how you would respond to claims of sen -- claims that social media

10 sensors right-of-center views, given that they are still a frequent topic of conversation,

11 still a political consideration, they're very much in the public discourse, and something

12 that you seem to have a pretty unique perspective on.

13 Do you -- do you have a reaction generally to those kinds of claims?

14 A I do. I understand those claims. And I believe that they are very much

15 rooted in a lack of transparency and accountability of content moderation practices and

16 policy enforcement and development decisions that happen behind closed doors by a

17 small number of people within social media companies.

18 I very much believe that that process is flawed and that there needs to, again, be

19 more transparency and accountability. That is part of why I am speaking to you today,

20 having been in those rooms and understanding the lack of insight that occurs and how

21 that lack of insight can then lead to the formation of a plan that, while not founded in

22 reality, makes sense when you don't have the reality to found them in, for instance, right?

23 And so that's very much how I think about these claims of censorship that are

24 happening specifically on right-wing views, right? I believe that there has not been

25 enough insight into the understanding of content moderation systems that would be able
 93

1 to say out loud, hey, we have research that very much shows that the abuse and toxicity

2 that is happening on these services also happens to align with the individuals who

3 espouse these views, right? And so, while it is not a political form of censorship, the

4 reality of who is and who is not doing abuse on services stands.

5 I would also say again, I think that these claims of censorship have become so

6 large and so overarching that I have seen decision-makers unwilling to make decisions

7 because they are paralyzed by the fear of being implicated and/or accused of having

8 some sort of bias against rightwing views. And that has led to, as I've said, logical leaks

9 and conclusions that don't make any sense in an attempt to evade those claims.

10 And so I think that these -- I think that the accusations of censorship have had a

11 tremendous impact in the exact opposite effect in that, while the accusations have been

12 over censorship of rightwing views, my experience has actually been a hands-off

13 approach to the highest level leaders of the rightwing movement based solely upon the

14 consideration that individuals do not want to seem biased and/or -- and/or politically

15 motivated in any way, shape, or form. Yes.

16 Q Thanks. I think that's really thoughtful. I'm glad we were able to get that

17 on the record.

18 I'll move on now to the lead-up to January 6th itself, then we'll talk about the day

19 of, and then we have some questions on the aftermath. But that should take us through

20 our last 2.5 hours here today. And I know it's a long day, so I'm really grateful for your

21 stamina and for your being here.

22 If we could pull up exhibit 6.

23 Thank you. And then, if we could -- it's -- the Bates number is 488.

24 If you give me a second, I can find the PDF page number for that. That should be PDF

25 page 39.
 94

1 So what we have here is what I understand to be meeting notes from at least the

2 safety policy team, maybe some other teams, dated -- it looks like these are dated in late

3 October. Is that -- am I right about that?

4 A Yes. These -- what I am looking at appears to be the note taken from a

5 safety policy meeting between both the U.S. team and the team in the Asia-Pacific region.

6 That's why there's two different dates, because during the time zone, it's actually two

7 different days at one time. So it was happening on the 22nd in Asia and on the 21st in

8 the United States of October 2020.

9 Q It's funny, I didn't think about that.

10 So during this meeting, it appears from the notes that your supervisor told the

11 safety policy team that it's, quote, North Star was to meet the reality of what is

12 happening on the ground in the United States and to mitigate harm both online and off.

13 Do you recall that remark?

14 I think if we scroll down a little bit, it's -- no, it's at the bottom of page 39. "Our

15 North Star," it's the last line.

16 A Yes. I am -- I am trying to recollect my memory, so let me read a little bit

17 here.

18 Yes. Do you mind repeating your question?

19 Q Well, my question was if you recalled that remark, but I -- I guess I was

20 wondering if you could describe the context of which that statement was made and what

21 it meant to you and your team?

22 A Yes. Okay. So October 2020 in a meeting between these two teams.

23 believe the context behind this was very much our teams, both in the United States and

24 around the world, were incredibly confused about the direction of our team, especially as

25 it was coming from leadership and our managers. And this, I think, can be reiterated
 95

1 from everything from around our -- our questions around and coded incitement to

2 violence to what you're seeing here, which is a very, very pointed and also very broad

3 question, right, which is essentially asking what is our goal point, right? What are we

4 aiming for? What are we trying to do here?

5 And it seems like our teams within Asia Pacific were specifically asking this

6 question of the leadership within the United States team, because very often our

7 Asia-Pacific teams they, as I was talking about with the dates, they picked up when the

8 United States went to sleep, right. And so they very much needed to have the context

9 and the information behind the decisions, because when we went to sleep, the rest of the

10 world was awake, and they were -- they were the individuals who had to deal with the

11 fallout. They were dealing with writing the assessments. They were the ones having

12 to deal with the questions.

13 And so I think at this point, our team had to go on through extensive leadership

14 changes, and we had a new manager who was leading things within the United States.

15 And we were all incredibly unclear about what our purview was, especially, again,

16 heading into an absolute historic election that we knew it was going to be historic. Our

17 team here was asking, what are we supposed to be doing? What would you like for us

18 to be doing and how should we be doing this?

19 Q Do you think that this -- I mean, this language is very potent, right. North

20 Star. Do you think that that was -- was that a consistent -- were they -- was your

21 management consistent in that -- in that goal, in pursuing that goal?

22 A Our management was not consistent in pursuing any goals. My team was

23 persistent in attempting to understand what the direction from leadership was. So this

24 question around North Star came up repeatedly within our conversations around the

25 world and with leadership. And they came up in relation to literally the team itself, as
 96

1 you see it's coming up here in relation to elections. It came up in relation to our hateful

2 conduct policies.

3 Our team was very, very much a group of highly intelligent individuals who were

4 very, very capable of understanding how to do our jobs and make recommendations but

5 were not being given decisions that made sense. And so we were lost. And so we

6 were very much asking to be given direction to do our jobs, right. In order for us to be

7 able to execute on this election and make sure that we are providing the safety that we

8 need going into it, what should we be doing? What should we be looking at?

9 Again, this North Star question, it came up repeatedly amongst our team and

10 specifically directed towards my manager, managers around the world, and, again, many

11 times directed towards Del Harvey herself and asking, what is not just the direction of the

12 policy team but the entire trust and safety team?

13 Q And there's a mention here of the -- the Proud Boys and their activity at this

14 time, sending threats by emails, of course, they were later on involved in several violent

15 protests. What -- what did it -- why did the team feel that it was difficult to defend

16 policies like glorification of violence in light of the Proud Boys activity?

17 A Let me read a little bit more here to understand. Is that in here down a

18 little bit later where it says something about glorification of violence?

19 Q Yeah. On the top of the next page, there's also some context.

20 A Thank you.

21 Yes. Yeah, I think -- I think this is very indicative of Twitter's policies at the time,

22 right? I'll literally read it out loud. For somebody on my team who was saying, it's

23 difficult to defend a policy such as glorification of violence -- which was in my team's

24 purview -- when we're seeing groups such as the Proud Boys using the platform to incite

25 violence. We need to take a better look at our policies and where we can make a
 97

1 heavier hand.

2 I believe that this was my team pointing out gaps within our policies and saying

3 out loud, there are -- and, going back to October, there are individuals that can be

4 identified as belonging to organizations that we know to be potentially violent that are

5 using our platform to hopefully stoke and incite their violence.

6 I believe at this time the Proud Boys may and/or may not have been officially

7 designated a extremist group, and it might have required them to be able to be

8 deemed -- again, my area was not in extremist organizations or in that group. But very

9 often there needed to be a designation for that to happen. This was the same thing

10 with QAnon, right? There needed to be a specific designation of that group as

11 something in order for us to take action on that.

12 And I think what we're saying here out loud -- what we're saying out loud here is,

13 there is a gap between what we know to be highly organized groups and organizations, so

14 the Proud Boys of the world or the Three Percenters of the world or all of those groups

15 that we know to be heavily organized around principles or whatever the case may

16 be -- but there is a gap between those individuals and what we see happening on the

17 service, which are individuals who are loosely affiliated with these folks, right, who might

18 share the same ideology, that we can't necessarily produce a membership card for but

19 are saying the exact sort of information that is going to lead to the incitement of violence,

20 and our policies do not have any way of taking that into account, and also my team

21 requesting for us to take a heavier hand in this sort of situation.

22 Q Does this link back to your previous conversation on offsite policy and your

23 ability to use offsite signals?

24 A Again, I don't -- we weren't thinking about off-service policies at this time.

25 And that's something that -- able to say, like, that's something that I thought of past
 98

1 since -- since this time, right, and was -- it was not -- it was not in my -- in my skill set at

2 that time to have really understood and acknowledged the kind of off-service policy.

3 But I do think -- I do think that in this sort of situation, we did have things like newspaper

4 articles, right, or things that were not happening on other services or platforms

5 necessarily that we needed to use to include as context, but we had information that was

6 saying that these individuals were involved in these sort of activities.

7 Q And what -- if -- on the previous page there was something about sending

8 threats by email. Do you know what that referred to?

9 A I do not remember, no.

10 Q Okay. So later in this meeting, your supervisor said that leadership was,

11 quote, comfortable taking risk about appearing less consistent.

12 I think I can find the full statement here. Someone asked, is leadership

13 comfortable taking this risk about appearing less consistent? And the response was that

14 yes, that they were.

15 What was the context of that question?

16 A Again, I think this is -- I think this is very much about trying to make sure that

17 our -- our various teams were in sync. Because I'm looking at one of these questions

18 that said something about market specific nuances, right. So this is -- this is a team

19 trying to be prepared, literally asking what -- what do we need to know to make sure that

20 we -- whatever is happening in your world we are fully understanding every single one of

21 the nuances when it lands on our plate and we have to deal with it.

22 Can you repeat that last question that you just asked about a specific thing, and I

23 want to look at that too?

24 Q Is leadership comfortable taking the risk of appearing less consistent? It's

25 difficult, especially from an APAC perspective, that there might be multiple different
 99

1 outcomes for one case.

2 Answer: Yes, leadership is comfortable taking this risk about appearing less

3 consistent.

4 A Yes, yes, yes. I think -- I think, again, that was -- that was the Asia-Pacific

5 team asking for backup and support, because if inconsistencies happen, they most likely

6 happen on their timeframe, right. Because whenever the decision was made, someone

7 would recognize it and say, well, this isn't what you all did last time. Why? And their

8 team was most likely dealt -- was tasked with responding to those sort of situations.

9 And, again, why would they want to understand the nuances in it.

10 And I think what they were asking, again, was as having to be the team that would

11 deal with the fallout, will leadership feel comfortable if and when we end up in a situation

12 in which these -- these disparate decisions that are being made, that we can't seem to

13 understand because there is no North Star, if and when we called out -- are called out for

14 being inconsistent, will leadership be okay with that? And according to what was said

15 here, they said yes, they would be okay with that.

16 Q So this is really useful because it establishes context to this document that

17 we didn't previously have.

18 I guess my last question before we move on is, if some of these same promises

19 about having a North Star of preventing on and offline harm and leadership that is

20 comfortable with a degree of risk enforcing policies that might not be public yet, would

21 that have also been of comfort to your team? Is that something you felt like your team

22 had or needed?

23 A If it was true, yes. But as you can see, I mean, leadership was literally

24 saying here that we're going to be taking the heavier hand, and this is in October. But

25 come November, when I was literally sitting in Del Harvey's office begging to have tweets
 100

1 removed that I was saying would literally lead to people shooting each other, I was told

2 no. So a heavier hand that is -- that is proposed or given here as the North Star in

3 October did not continue to be the North Star come 3 weeks later.

4 Q And Del Harvey was not in this meeting, or was she?

5 A I do not believe Del Harvey would have been in that meeting.

6 Q And so the person leading this meeting would have been your direct

7 supervisor?

8 A I believe so.

9 Q Okay. Did you ever request this kind of feedback or have this kind of

10 conversation of the sort that you had with Del Harvey about clarifying these policies with

11 your -- with your supervisor?

12 A We had many conversations with my supervisor attempting to clarify our

13 policies.

14 Q How did those go?

15 A Not well. There were -- there were several of them. So I do not

16 remember any of them going particularly well. If there are specific ones that you would

17 like to ask about, we can do that.

18 Q Well, let's see, if we can move to -- I think it's pages 10 and 11 in the PDF of

19 this document. I'm not sure we have specific meetings, but I think we can talk generally

20 about whether or not you got the type of guidance and feedback and upward

21 management that you might have liked.

22 A Yeah. The answer is no. But, yes, we can definitely -- we can definitely

23 talk through it. I did not get the support or guidance in management that I needed.

24 But yeah. Sorry to cut you off.

25 Q So I'm looking at this document now, and this appears, I think, to be

 101

1 feedback for your supervisor, right?

2 A Yes.

3 Q So some of the things you flag in this document are lack of communication.

4 You say that your supervisor frequently did not attend team meetings?

5 A Yes.

6 Q How did that work? I mean, how did that affect the work of your team?

7 A It was -- I believe the impact of that is what you just saw on those team

8 notes, right. The confusion, the utter chaos of trying to understand what direction we

9 should be pointing towards while also trying to move full-steam ahead at the same time.

10 Yeah.

11 Q I guess another related question is, would it have been normal for you to be

12 having conversations on this with the vice president, or was that a result of inadequate

13 guidance from your direct supervisor?

14 A I can't say that anything at this time was normal. And I'll say that because,

15 as you can see from this document, my supervisor -- my new supervisor had been in their

16 role for 100 days -- only 100 days at this time. And as you can also see in there, I say I

17 have had three managers within 1 year. And so at this specific moment, the former

18 global head of our team had actually been terminated for creating a psychologically

19 unsafe work environment for us.

20 And so this manager that I was writing this feedback on had actually been hired

21 specifically only to manage people, not to manage policies, and they were not hired to

22 manage me. And yet, given the turn of circumstances and an individual being

23 terminated, we lost a layer, we no longer had a global head that was responsible for

24 policy decisions. And, instead, there was the power vacuum that existed in which a

25 couple of individuals, including my manager, attempted to try and fill in to make

 102

1 decisions. And, again, it was incredibly unclear who was the actual decision-maker.

2 One of the individuals who was also in that sort of power vacuum was the

3 individual that I sent that message to specifically around January 4th through 6th trying to

4 make those decisions.

5 So, at this time, there was chaos. And so I don't know if it would have been

6 normal or not normal for me to have been in conversation with the vice president about

7 these specific issues. I was the most senior member of the United States team, and so it

8 would have fallen on me, especially given my subject-matter expertise, to represent the

9 United States team to the vice president within these circumstances. And we did not

10 have a manager who had any tech policy experience, who had any free expression

11 experience, who had any policy experience, who would have been able to make these

12 concrete decisions or to be able to represent the opinions or the ideas or assessments of

13 our team in any sort of appropriate way. So it was all over the place.

14 [Discussion off the record.]

15 BY
16 Q Okay. So, yes, this document says that this -- your manager had been in

17 the role for 100 days. I think it's undated. Do you remember about what time you

18 wrote this feedback?

19 A It had to be winter of 2020. It was before January 6th.

 103

2 [3:54 p.m.]

3 BY

4 Q Potentially after the election?

5 A I think it was after the election, yes.

6 Q Okay. And so --

7 A 1'm pretty sure it was after the election.

8 Q And so your team had really made it through a tense period with a lot of

9 management changes and then a manager who had only been around for about 3 months

10 and was not trained in the actual subject matter of your work.

11 A Yes.

12 Q And concerning the global head of policy position, was there an intention to

13 eventually fill that, or were they trying a new structure?

14 A They were hoping to fill that position. I believe it has been filled now. But

15 at the time when I was there, there was no one in that position.

16 Q So this position, would it have normally been the direct go-between for you

17 and Del Harvey?

18 A Yes.

19 Q Okay. So it's a pretty important role at Twitter.

20 A Yes. It is a hugely significant role that has immense amounts of power and

21 responsibility that impacts not just the Twitter service but the daily lives of all of

22 humanity.

23 Q That's -- I mean, that's sobering.

24 When did the previous person leave this position? How long had the position

25 been vacant?
 104

1 A I don't know when they had been hired. I believe they were hired in maybe

2 2018 or 2019 and then was fired in 2020.

3 Q Do you remember, was it, like, early 2020? Was it the summer?

4 A It was the summer of 2020.

5 Q Okay.

6 And what role -- so you described this person as being incredibly important for the

7 day-to-day political discourse of humanity. Like, walk me through, like, in practice, how

8 they play that role. What did they -- how did they --

9 A Yeah.

10 Q -- what role did they play in managing the platform?

11 A Yeah. Yeah. So the individual who would've been global head or regional

12 head of the site policy team was the first point of decision making for all of our teams.

13 Again, we were recommendation-makers.

14 And so they were responsible for not just, again, the safety and privacy but, as

15 well, the terrorist organizations, violent extremist groups, and the child exploitation

16 teams as well. And so any escalation that would come into the various policy teams all

17 around the world from these areas would go to that individual in order to make the first

18 gut check or first decision. Very often, that person was the last line of decision making if

19 something needed to be happening very, very quickly.

20 And if that person could not make the decision, they were then responsible for

21 then going to the vice president and representing the opinions and the thoughts and the

22 recommendations that came specifically from our team to the vice president in order to

23 make those sorts of decisions.

24 Why I say it's so important and it's so vital is very much, again, because the team

25 that I sat on was responsible for the safety of the service and what happened to
 105

1 individuals not just on Twitter but literally their bodily harm and their lives and what

2 happened to them as an implication of the speech that occurred on our service. And

3 those decisions could literally be life or death for individuals. They could also be foreign

4 policy and implementation of huge political decisions and discourse around the world

5 that could change things. Yes.

6 Q Thank you. That's a really valuable perspective for understanding a role

7 that was vacant at this time and struggling to be filled by someone who really was not

8 framed to fill it, and so you've had to step in and interface directly with the vice president.

9 A Yes.

10 Q You later write in this document that "leadership did not have an adequate

11 understanding of the scope of the coded incitement project and were confused --

12 A Yes.

13 Q -- about its origin, urgency, and ultimate purpose."

14 A Yes.

15 Q That's a very astute summary, I think, of our conversation so far.

16 When you say "leadership," I guess walk me through who you mean and --

17 A Yes.

18 Q -- what their various stances were or where their various shortcomings were.

19 A Yeah. I think that they're referring to Del Harvey in this conversation and

20 they're referring to the conversations that I had had with Del around coded incitement to

21 violence in which Del was taking a very nonchalant approach, as we have discussed, right,

22 coming in with very interesting self-defense theories and an unwillingness to act.

23 And, from my perspective, again, we were tasked with a sense of urgency. And

24 so it seemed clear to me that the urgency that our team and the directive that we had

25 been given to very quickly solve something that was rapidly becoming a problem had not
 106

1 been reiterated to Del.

2 And so it made me wonder what the conversation that had occurred between my

3 manager and Del was and whether, again, the opinions and assessments and

4 recommendations of our team were adequately being represented to Del in such a way

5 that would have underscored why the coded incitement to violence policy was essential

6 to be shipped and launched back in November, rather than be put on ice and move

7 forward in a sense of confusion.

8 Q Just thinking about all of your interactions with Del, do you think that it was

9 that your concerns weren't being translated up through your supervisor, or was it that Del

10 had a different point of view or a different practical stance?

11 A All of the above.

12 Q All of the above. Do you think that Del's position -- would you characterize

13 her position on the coded incitement issue at this time as reckless?

14 A Yes.

15 Q I know we've already discussed the December 19th "be there, will be wild"

16 tweet. I'm wondering if you could talk more about how your supervisor and Del reacted

17 to that tweet and the change in political tenor on Twitter after that time.

18 A I don't know that they did react.

19 Q No reaction?

20 A And, again, that was, for me, cause for concern, that there seemed to be

21 no -- there seemed to be no urgency, there seemed to be no emergency, there seemed to

22 be no crisis. Which, again, was the impression that I was very much under and that we

23 were told to be under by other individuals.

24 And yet, when it came to our team's leadership, when the leader of -- the

25 President of the United States is tweeting to his followers, who are becoming increasingly
 107

1 more armed and increasingly more violent and expressing their willingness to participate

2 in violence, and is giving them not only a place but a location to do such and enact such

3 violence, I was increasingly concerned and flabbergasted that the individuals whose job it

4 was to actually be making the decisions and implementing policies that would have us be

5 prepared for these circumstances were doing literally nothing.

6 Q And, during this time, to reiterate, your supervisor was not attending

7 meetings, kind of absent from the policy discussion.

8 A Yes. There were many, many times that they were late to meetings.

9 When they did come to meetings, they were on their computer, messaging people, or

10 messaging the whole time, not paying attention.

11 There were meetings that were so contentious that they actually asked me to

12 start attending meetings so that I could attempt to navigate the contentiousness better,

13 because I had a good relationship with everybody on the team and I would've been able

14 to step in and attempt to do some sort of mediation in those circumstances.

15 Yes. It was -- it was quite a time.

16 Q And these contentious meetings, would they have been around, like, the

17 potential for violence in the election? Is that what they were --

18 A Yes. I mean, it --

19 Q -- the contention was?

20 A It was around for potential for violence in the election. It was also around

21 things like the North Star, right?

22 Our manager at this time, again, with no policy -- no fault of their own, does not

23 have the experience or the background to be able to do this work, right? And so, when

24 you have highly skilled individuals who have been doing this for an incredibly long time

25 who are asking for very, very simple things like what direction should we be heading into,
 108

1 and their manager can't provide that sort of solution, it becomes incredibly contentious.

2 Because we're just trying to do our job, and we're not able to do that because our

3 leadership is literally standing in the way and disallowing us from taking actions that

4 would allow us to do that.

5 Q And given that you were one of the strongest advocates for taking stronger

6 action, that must have been an uncomfortable position, to be asked to moderate that

7 debate?

8 A It was not -- it was not my favorite time.

9 And, again, the work that the safety policy team at Twitter does is so essential to

10 the functioning and survival and protection of humanity that I recognized the absolute

11 dire need for this team to at least in some sort of way, form, or fashion move forward and

12 move along. And so I was willing to show up to meetings and to be that intermediary, to

13 attempt to get those decisions, to literally keep raising questions until we were able to

14 get those answers.

15 Because I was also one of the most tenured people on the team, and so I had been

16 around long enough to understand what happens when we did nothing, right, or what

17 happens when decisions were not made and how important it was to have decisions

18 made.

19 And so, yes, it was -- while I would not like to go back to that time in my life, it -- I

20 knew how important it was for that team to function and how important it was for that

21 team to do its job, and so I -- that's what I was trying to do.

22 Q And how was your feedback received by Twitter? You provided this

23 360-degree feedback. What did Twitter do with it?

24 A So this feedback went to both my manager as well as Del Harvey.

25 I don't know if we specifically talked about it, but one of the recommendations
 109

1 that I made within this feedback was that my manager be retrained on the policies that

2 our team worked on, because I believed and had experienced their background

3 information and skills to be incredibly lacking in the area.

4 And specifically what I said -- I think if you scroll down in here, I pointed to coded

5 incitement to violence as an example of a policy that, if this individual may have had

6 training and understood our original violence policy -- so the policies that were actually

7 on the books -- maybe then they would've been able to have a little bit more of a deeper

8 appreciation or ability to understand the nuance of something like a coded incitement to

9 violence policy.

10 Q And it looks like you say here also to be a better advocate for you and your

11 team and your positions.

12 A Yes, very much so. Very much so.

13 Q Did you ever get any indication that Twitter was going to follow through with

14 that recommendation?

15 A I did not.

16 I mean, if you see, one of the last things I say on here is that I would like to have

17 this person "help address the open questions that remain to this day about our approach

18 to this content," which was specifically coded incitement to violence, as I say up there,

19 the incredibly complex coded incitement to violence policy.

20 No, actually, after I submitted this feedback, my manager began to treat me

21 negatively, I believe based upon me giving information to their boss that they might not

22 have been as knowledgeable or as experienced or as skilled as they needed to have been

23 to be operating in the position that they were in.

24 Q Thank you.

25 I wonder if we could move to page 9 of the PDF document.

 110

1 So this is another -- it looks like potentially a Slack message. There's not a lot of

2 context to this document. It's dated February 11th. I wonder if you recognize it and

3 you could just provide some details about what we're looking at.

4 A Yes, I do recognize this. So this is a -- what are they called -- a note, on a

5 MacBook. And the language included within this note is information that I submitted to

6 one of Twitter's research teams that -- I submitted it to one of Twitter's research teams.

7 I did that because this research team had sent out a survey to the entire company

8 essentially asking what, if any, part Twitter played or had to play within the events of

9 January 6th. Yes.

10 Q So was that team then tasked with writing any kind of retrospective account

11 of Twitter's role on the 6th?

12 A I am not 100 percent sure what the entire purview of that team was.

13 However, it was ran by somebody that I knew to be a very good researcher and someone

14 I knew to do very good work. And so I wanted them to have --1 wanted them to have

15 the true information in order to be able to do their good work. Again, I don't know what

16 the exact work product was for this.

17 I will also say that, while I did know the individual who was working on this, I also

18 reached out to them to specifically ask if the feedback would be anonymous, because I

19 was very fearful that if I were to give this feedback and say explicitly the truth and what

20 had happened that I would be in trouble.

21 Because the tenor that I was receiving from the research team, again, was:

22 What, if anything, did we -- did leadership know anything, and if so, like, how little did we

23 know? Right? It was a very, very hands-off approach and the complete opposite of

24 what I knew to be true, which was: For months, we had been seeing this content,

25 begging for it to be removed. We had been specifically raising it to leadership. They

 111

1 had specifically told us we could do nothing about it until violence actually broke out and

2 then we were asked to clean it up.

3 Q To your knowledge, was any retrospective -- were there any other

4 retrospectives that Twitter conducted?

5 A Yes. There was -- I believe there was a retrospective being conducted by

6 outside counsel in --

7 Ms. Ronickher. Well, let's not talk about the specifics.

8 The Witness. I will not talk about the specifics.

9 BY
10 Q That's fine.

11 So outside counsel ran that retrospective. This question might seem somewhat

12 abstract, but if you'll just entertain it. Is a retrospective like that something you

13 personally believed would be important for the new global lead for site policy to

14 familiarize themselves with?

15 A Absolutely.

16 Q Would you expect a new person coming into that role to read that

17 document?

18 A Absolutely. I don't think -- I think somebody coming into that role as the

19 new global policy head absolutely should be able -- should and needed to have some sort

20 of retrospective or information about these events and these days.

21 I asked several times for a retrospective to be completed, specifically around the

22 days of January 6th through January 8th. I asked this because, again, as a person who

23 firmly believes in transparency and accountability, having been an individual who had

24 been involved in the process, I believed very much that those recommendations and

25 those decisions needed to be analyzed by individuals outside of just ourselves and be able
 112

1 to be understood within context.

2 I also recognized, as an expert and scholar of the field, that these decisions and

3 these actions that were taken during these weeks were the most important and biggest

4 content moderation decisions of history up until that point, of modern history, and that

5 doing them without any sort of retrospective or doing them without having any sort of

6 reflection back would be doing a disservice to the field that I worked in and that I have

7 studied and that I have deep respect for, because of the reality of, again, knowing how

8 big and how important and the realities of the industry and how many people would sue

9 based on this and knowing that we needed a record and all of these things.

10 It was -- it was -- it still blows my mind that that did not happen.

11 Q - h a s an immediate followup.

12 Go ahead.

13 BY
14 Q Yes. Thank you, J. Smith.

15 Going back to the point about the global head of site policy, can you explain

16 exactly why it is important for that individual to have maximum familiarity with the

17 decisions that were made around Twitter's response to January 6th even though that

18 event is, obviously, in the past?

19 A Yeah. Well, I think -- I think the biggest reality is that, while this event

20 might be in the past, it is very much also a part of our future. One of the reasons why

21 I'm here and why I'm speaking to you is because, from what I see and from what I know, I

22 very much believe that January 6th is going to happen again and it is going to be worse

23 and it is going to be bigger.

24 And so I very much believe that, in order to attempt to -- or make some sort of

25 intervention in which that reality is not true and that prediction does not come true, we
 113

1 need to have a familiarity and an understanding of what actually happened during that

2 time in such a way that we can understand what went right, what went wrong, and how

3 we can either fix and improve upon those situations or not literally replicate the exact

4 same mistakes.

5 And that has not happened. Up until this point, the individual who is going to be

6 running the head -- the global head of site policy will be facing any other protest

7 situation, will be going into the 2022 elections, will be going into the 2024 elections

8 without the reality and without the understanding of what occurred between January 6th

9 through January 8th and, again, the mistakes that were made or the things that were

10 done well there that could be done in order to anticipate the same issues that we know

11 are already developing for these future elections.

12 Q Thank you.

13 And one more question. Would you consider the head of site policy someone

14 who's in the need-to-know category for all information related to Twitter's response to

15 January 6th, given incidents of far-right violence that we've already seen in the

16 aftermath?

17 A Yes.

18 Q Okay. Thank you.

19 A Yes, I would absolutely believe them to be a person that needs to know.

20 Q Thank you.

21

22 Thanks,

23 Thanks, J. Smith.

24 I want to bring us to the day of January 5th. I think the --

25 Ms. Ronickher. Sorry, I'm just going to jump in and check in with my
 114

1 client to see if --

2 Yeah, of course.

3 Ms. Ronickher. We've been about an hour, a little over an hour. Are you doing

4 okay? Do you need a break?

5 The Witness. Let me take, like, two sips of coffee and then --

6 Ms. Ronickher. What about a 5-minute break?

7 The Witness. Let's do it. Let's do 5 minutes.

8 We can do 5 minutes. Yeah, of course.

9 The Witness. I like that.

10 Let's close the record at 4:20 p.m. and reopen at 4:25.

11 The Witness. Perfect.

12 Thank you, Alexis.

13 [Recess.]

14 The record can reopen at 4:25.

15 We've got about an hour and a half left and the two most important days to cover,

16 as well as some of the aftermath. So we'll try to make pretty good pace through this

17 portion of the questioning.

18 BY
19 Q On January 5th, it is the select committee's understanding that there was a

20 staff meeting among the safety policy team at which many members of the team

21 expressed concerns about the possibility of violence on the 6th.

22 I want to ask for the record, was your supervisor present at that meeting?

23 A No, they were not present at that meeting.

24 Q And the person who did run the meeting is based on European time and

25 would've been, sort of, the next most senior manager. Is that right?
 115

1 A Yes. She works on the European team and is the -- I believe the director of

2 that team.

3 Q Okay.

4 Now, could you describe, maybe, some of the concerns that were raised during

5 that meeting and some of the questions you asked that manager?

6 A Yeah, absolutely.

7 So this meeting was a regularly scheduled, I believe, safety policy team meeting

8 that -- we recorded all of our team meetings because our -- this specific meeting was

9 happening between, I believe, the European office and the U.S. office. As you saw in the

10 team notes, we have other meetings that happen, specifically between the Asia-Pacific

11 office and the U.S. office. So we recorded our team meetings such that other teams

12 could all watch them.

13 So, on January 5th, we had a meeting, and I specifically made an agenda point

14 item asking, for one of many, many times, if we could clarify our approach to coded

15 incitement to violence. Specifically, as we knew that individuals were going to be

16 convening on D.C. the next day, I was hoping and urgently pleading that we have some

17 course of action in place in anticipation of the violence that myself and many others in

18 the meeting also raised concern that was imminent to happening the next day.

19 Q And what was the response that you received?

20 A It was a lot of the same response that I had received from leadership and

21 from management on the team. I believe this individual reiterated Del Harvey's

22 example of self-defense mechanisms and not wanting to remove people who were saying

23 they were locked and loaded in case they were in their houses waiting for self-protection.

24 So it was a lot of regurgitation of the same nonsensical arguments that had been made

25 previously to that meeting.

 116

1 We were also, again, told that we were not to be taking the material down.

2 Q Why was it that, after 2 months, I mean, "locked and loaded" still seems to

3 occupy such an outsize role in these discussions? There were other phrases that were

4 researched and frequently used, and yet the rebuttal to your urgent pleas always seems

5 to be about "locked and loaded."

6 A I'm not sure, to be honest with you. I think that that might have just been

7 a tidbit that came up that was, again, being regurgitated. I think "locked and loaded"

8 was one of the most highly contentious phrases that was happening in all of those that

9 we saw, because, again, it was specifically relating to the usage of arms and, again,

10 speaking about guns and things that could be used to instill violence.

11 Q But, I mean, you were also not allowed to action content using "stand back

12 and stand by," for example --

13 A No, we were not.

14 Q -- even though that was the phrase that initiated the need for a coded

15 incitement policy.

16 A Yes, it was. And we were not allowed to action it, correct.

17 Q And so you are in this meeting. You ask the manager on duty if this policy

18 can be used in the event of a crisis. They say no.

19 And then this person is on European time. So what happens when they go to

20 bed, when they're unavailable?

21 A So I believe -- I had requested to use -- I was requesting a lot of things.

22 believe I was really -- again, it was the day before, and so I was attempting to find any sort

23 of solution that we could piece together to have us prepared. And so I was asking

24 about -- again, we were not decision makers. And so I was asking who was on point to

25 be our decisionmaker during these times and who we should go to in order to get those
 117

1 decisions made quickly.

2 This individual said that she would be available -- they would be available until

3 they went to sleep in European time, and when they went to sleep, we should ask a

4 manager, who I believe had been at the company for maybe a month and had not been

5 trained yet on any of the policies. She recommended that we ask that individual for

6 signoff in order to action.

7 Q So this new manager, this now, kind of, third in command, if you will, what's

8 the time window during which they would've been the point for escalation?

9 A The European team went offline around 11:00/noon East Coast time.

10 Q So they would've been on around noon East Coast time on the 6th.

11 A Yes. Yes. Yes.

12 Q And that's who was on duty, then, for the breach of the Capitol Building.

13 A Yes.

14 I also believe my --1 don't believe; I know -- my manager was also -- my manager

15 had experienced a loss in the family and was back on January 6th, was not available on

16 January 5th.

17 Q I see. Okay.

18 So you have issued this explicit warning in a meeting and asked for authorization

19 to use a policy that's already been prepared but not allowed to be implemented. You've

20 been told to make due, essentially, with what you have and that you'll react to events

21 as they occur, if they occur.

22 I mean, that night, how did you feel? What was your gut feeling on the night of

23 January 5th?

24 A I mean, I believe I even said in the meeting something along the lines of -- I

25 mean, I'm trained as a lawyer, and so I went back to my training and said, like, "I want to
 118

1 say on the record that I am very concerned about what is happening on our service and

2 the action-slash-inaction that we have taken for months." And I also expressed my

3 concern that violence would occur the next day and that, when a violence occurred, there

4 would be this content on the service. And so I said that in the meeting.

5 I believe I sent a Slack message to someone that said something along the lines of,

6 "When people are shooting each other tomorrow, I will try and rest in the knowledge that

7 we tried."

8 And so I went to -- I don't know that I slept that night, to be honest with you.

9 was on pins and needles. Because, again, for months, I had been begging and

10 anticipating and attempting to raise the reality that, if we made no intervention into what

11 I saw occurring, people were going to die. And on January 5th, I realized no intervention

12 was coming, and even as hard as I had tried to create one or implement one, there was

13 nothing, and we were at the whims and the mercy of a violent crowd that was locked and

14 loaded.

15 Q And, just for the record, this was content that was echoing statements by

16 the former President but also Proud Boys and other known violent extremist groups?

17 A Yes. I think I remember, even in the meeting, one of our colleagues looked

18 on Twitter at that very moment and saw that there was content, literally at the moment

19 that we were in our meeting, that was problematic, and we weren't allowed to take it

20 down at that time either.

21 Q Okay.

22 If we could pull up exhibit 6 again. It's page 1 of exhibit 6.

23 Ah. Here we go.

24 So we'll turn to these messages in a moment, but, earlier, I think way back at the

25 very beginning of our conversation, you said that when Twitter integrated -- that Twitter's
 119

1 enforcement terms on wish of harm and violent speech changed over the course of the

2 2 months between the election and 1/6 and that it changed to allow it to issue strikes and

3 ultimately to suspend users.

4 Why -- was this change -- were any of these changes made on the morning of

5 January 6th?

6 A Yes. Yes. So I can tell you -- do you want me to tell you a little bit about

7 that?

8 Q Yeah. And, to clarify, I think I asked this at the time, too, but this was the

9 morning of January 6th East Coast time?

10 A Yes. So, morning of January 6th East Coast time, I got to my computer. I,

11 again, was incredibly anxious and logged in to our team Slack channel. And I saw that at

12 some point over the night my manager had made the decision to permanently suspend a

13 user underneath the coded incitement to violence policy.

14 And, when I saw that, I was incredibly confused, one, because, again, we were told

15 we were not allowed to use the policy and that it didn't exist, and so, now, on the

16 morning of January 6th, it not only existed, it was being used, and not only was it being

17 used, but it was being used at the most punitive level that existed, at the point of

18 permanent suspension, which was nothing that we had ever spoken about.

19 I was also incredibly concerned because the account that was permanently

20 suspended, when I looked at the content that was in question, I believe it was me mes.

21 It was jokes and satire material that they made the decision to permanently suspend

22 over.

23 And, for me, I was highly confused because it was nowhere near as egregious as

24 the content that I had been, again, asking to take down for months. And yet, the

25 morning of January 6th, individuals were being permanently suspended for a

 120

1 non-egregious joke and satirical content under a policy that I was told I wasn't allowed to

2 use.

3 Q So you weren't really referring to a conscious, thoughtful policy change in

4 reaction to anything. It was a surprise, kind of, one-off incident that you felt was

5 actually in violation of standing policy.

6 A Absolutely. I was shocked.

7 Q That's helpful. That's really helpful. That contextualizes a lot of things for

8 me.

9 A Uh-huh.

10 Q So the messages I've pulled up here, again, I don't think we have exact dates

11 on these, but I, kind of, from context, believe they're from January 6th?

12 A Yes. I believe they are as well.

13 Q And it seems like you're saying that your supervisor, first off, was able to

14 report to work that day after the tragedy in their family, but then made this decision, and

15 you were frustrated that they had done this early in the morning before going to the gym.

16 Is that right?

17 A Yes. So I actually think that this happened later in the day. So I believe

18 that the Slack conversation that I'm looking at right now occurred -- it says 2:18 p.m. So

19 I believe that this actually occurred a little bit later in the afternoon, after I had had a

20 conversation with (REDACTED) -- my former supervisor and another team member.

21 Yes. I'm happy to talk about that conversation and kind of move backwards, but

22 this is a little further in the day.

23 Q Yeah.

24 And for the court reporter, if it's possible to redact the name that

25 was just said. We can also take care of that later. But I recognize that desire.
 121

1 BY
2 Q So these messages at 2:18, I assume Pacific time, were probably, then, after

3 the attack?

4 A I be- -- yes. Yes. These messages occurred after the Capitol had been

5 breached and I had, again, had a conversation with my supervisor about what we should

6 do and how we should move forward with taking action, given the violence that was

7 occurring on the ground.

8 Q Okay.

9 It's our understanding, both from these documents and from the timeline that

10 Twitter provided, which you viewed earlier, that, ultimately on January 6th, at the time of

11 the attack itself, as the Capitol was being stormed, right, Congress is being evacuated,

12 certification of the election has been postponed, Twitter does begin to action tweets for

13 coded incitement to violence.

14 Is that correct?

15 A Yes, but -- yes. Yes, and it took a little bit chime, so it wasn't -- it was a

16 little bit delayed. Yes.

17 Q Let's walk through the events of the day.

18 A Okay.

19 Q So how was that morning to early afternoon for you before the attack?

20 How did your team spend their time? What were you noticing? What were you

21 doing?

22 A Yeah. That's a great question.

23 So, after I woke up and saw that the individual had been permanently suspended

24 for coded incitement to violence, I specifically asked my manager -- again, you see here

25 the word "threshold." I actually haven't said it here, but I wouldn't stop asking at the
 122

1 time, what is the threshold? So at what threshold can we enforce for coded incitement?

2 What do we need to see? How far does someone need to go in order for it to happen?

3 And so I asked them, after seeing the permanent suspension, again, what our

4 threshold was. And I made it very clear, we were going to need to have an answer to

5 that.

6 They said that they would be meeting with the larger leadership team, including

7 Del, later in the day. I asked if I could attend that meeting. I was not invited to attend

8 that meeting.

9 I will also say, I did send -- I was so concerned about what was going to happen

10 that day that I actually sent a message expressing my concern to our lawyers.

11 Alexis, am I allowed to talk about the details of that?

12 Ms. Ronickher. You can say what your -- actually, just to be safe, we should

13 leave it at that.

14 The Witness. Fantastic. Okay.

15 I -- because I was so concerned I believed it to be a legal liability, I went as far as

16 reaching out to that team with my concerns in an attempt to then express those

17 concerns. So that happened the morning of January 6th.

18 Later in the afternoon, I believe the meeting that happened with leadership

19 happened around 1:30, I want to say, EST.

20 After that meeting occurred, I was then called into a meeting with my supervisor

21 in addition to another member of the safety policy team who was the other senior

22 member of the team. And, at that time, my manager approached us with two specific

23 tasks. First, we were tasked with stopping the insurrection and, then, finding a way to

24 permanently suspend the former President's Twitter account.

25 BY
 123

1 Q And so this directive to stop the insurrection, were you given a sort of green

2 light to, you know, yes, use coded incitement? Or was it -- was it a -- were you sort of

3 forced to act on your own initiative?

4 A I wish that I would've been given the directive and, again, been -- have

5 leadership that was able to understand and have the comprehensive nature to be able to

6 help establish that threshold that was needed.

7 Unfortunately, what actually occurred was, at that time when I was asking -- and,

8 again, we were given the directive, and I said okay. What I actually did was, first, I made

9 sure that I said out loud, "I want to express my frustration with you," to my manager.

10 And I said to them, "I told you that this was going to happen. You did not listen to me,

11 and now you are asking me to clean it up." And I told them how frustrating of a

12 situation that was.

13 And then I told them, in order to do that, I needed to understand what the

14 threshold that I had been asking for for months was. They could not -- they couldn't

15 give me a clear answer.

16 And so what I did is -- there was a tweet that had been escalated to us that our

17 team had found to be in no violation. And a member of our legal team escalated to

18 me -- Alexis, am I allowed to say what their recommendation was?

19 Ms. Ronickher. [Nonverbal response.]

20 The Witness. No.

21 We were asked to re-review.

22 And, at that time, I -- it was in this meeting with my manager, and I sent them this

23 specific tweet, and I said out loud, you know, "Our team had assessed this this way.

24 There is another assessment that could go this way."

25 Let me say just here very quickly, that is not out of the ordinary. Almost
 124

1 everything that landed on our desk could go one way or the other.

2 And so what I was saying in this specific circumstance when we were looking at

3 this, "Is this a violation or not a violation, Manager?" And they shrugged their shoulders

4 and said, "Eh, I don't know."

5 And so, at that moment, recognizing again that the tweet that I was looking at was

6 actually a picture of a barricade that had been knocked down and was announcing a

7 breach in the location at the Capitol Building and was calling on individuals to come to

8 that area in order to access and breach the Capitol, I looked at my manager and I said,

9 "We need to make decisions. It is time to make decisions."

10 And I said, "You are not able to make decisions at this moment, and so I am going

11 to have to make the decision, based on this other advice that I was given, that this is

12 going to be the threshold in which we are going to start taking down content." They

13 nodded.

14 And so, from that moment, I then began to say and formulate, kind of, our policy

15 approach by saying, we are going to err on the side of caution in understanding where

16 our threshold was.

17 I can talk a little bit more about it, but then I -- there was a little bit more that

18 happened in that meeting, but after I left that meeting, I went to go work on the

19 enforcement guidance to do that.

20 But I'll pause here.

21 BY

22 Q And when you left that meeting, about what time was it, ballpark?

23 A I'm assuming, based on these messages, somewhere around 2:20 p.m.

24 Q So, around 5 o'clock Washington time. I think around then there were

25 signs that the Capitol was at least going to start to be cleared --

 125

1 A Yes.

2 Q -- the worst of the violence had been done. And you were unable to

3 receive clear guidance from your manager on instructions on how to breach Federal

4 property during a constitutional process?

5 A Yes. They shrugged their shoulders at me.

6 Q And so, really, this is an abdication of responsibility on the part of Twitter

7 leadership, which left you to bear the full brunt of their day-of, moment-of

8 decisionmaking?

9 A Absolutely. That is exactly what happened.

10 Q You may not be able to answer this question, but, just revisiting the message

11 you sent to Twitter's legal team before the January 6th attack, are you able to

12 characterize why you thought it was a liability? Like, just what the -- like, what the class

13 of liability might have been? Or is that too -- is that too intrusive?

14 Ms. Ronickher. Nope. I think that's fine. She can talk about her thinking.

15 The Witness. Yeah. I did not want the blow-back to come back on my team,

16 because we had tried our hardest. And it was very clear to me that we did not stand a

17 chance and we were in no way prepared for what was about to happen and what was

18 coming our way that day.

19 And knowing how many, again, lawsuits stem out of decisions, enforcement

20 decisions, that are made on Twitter, I recognized that as a day that would be

21 astronomical for that sort of liability, and, again, wanting to say out loud: My team

22 tried, and this is on -- this is on leadership that very much would not allow us to make

23 those decisions.

24 Thank you very much.

25 It's almost 5 o'clock, and I want to give a chance to ask any followup
 126

1 questions. And he'll also continue this line of questioning. We'll probably return to

2 that timeline document and see how Twitter characterized this course of events, but,

3 -
4 - Yes. Thank you,

5 So I think what we can do -- I think I'll certainly have followup on everything you

6 were discussing, but maybe the best way to do it is to continue working through our

7 outline, and then we can sort of deal with followups as they come up.

8 So, if we could bring up exhibit 3.

9 Great.

10 BY-

11 Q And we actually talked about this before, J. Smith, but I wanted to go back to

12 the moment of 2:20 p.m. eastern time and clarify for the record that I think -- I'm not sure

13 if those messages we were just talking about were on eastern time or on Pacific time.

14 Would you have any insight into that?

15 A I do not officially remember, no.

16 Q Well, regardless, it would have been after the initial breach of the Capitol.

17 And if, assuming, again, arguendo, this timeline is precisely correct and that it was

18 eastern time at 2:20 p.m. when Twitter started to identify and review coded

19 language -- we talked about this earlier, but I wanted to ask about whether your team

20 was at all responsible for enforcing this change in posture once that had been made.

21 A Yeah, absolutely. Absolutely.

22 So this moment of identifying and reviewing the coded language came after the

23 meeting I had with (REDACTED) where they absconded their decision making power and I

24 recognized again that people's lives were hanging in the balance and a decision needed to

25 be made.
 127

1 From that, I left the meeting and I started drafting enforcement guidance for

2 January 6th based specifically on the coded incitement to violence policy that was written

3 previously. I relied heavily upon that document in order to create and understand what

4 we should be doing with content on that day. And so I used the policy language and, I

5 believe, the recommendation for abuse one-off as the action.

6 And I worked with two other members of the safety policy team to create that

7 document that outlined how we would be implementing coded incitement to violence

8 that day, on January 6th, in specific relation to certain content and materials that we saw

9 on the service. We collected tweets as examples, inserted that into the document.

10 That document was reviewed by legal. I believe it was also sent to Del Harvey.

11 And then it was shipped, and by "shipped" I mean emailed and sent to content

12 moderation teams all the around the world as well as Twitter Services. This included

13 what we called a resource request, which essentially says: We need everybody,

14 anybody who is available, we need you specifically on this resource -- on this request,

15 looking specifically for this content.

16 And that takes time, right? Again, it takes time for the email to be sent; it takes

17 time for moderators around the world to read the document, to understand what's going

18 on. And so, in the meantime, myself and my team took a highly unusual action that we

19 typically never did, but we hopped into Profile Viewer and began also searching and

20 finding this content ourselves and bouncing it on the back end ourselves.

21 Q Thank you for that timeline.

22 And so, if you remember, about how long did it take your draft document to get

23 approved and shipped out?

24 A It was a very quick turnaround. Less than 2 hours.

25 Q And to revisit earlier our conversation, this was the same kind of policy you
 128

1 had been urging for Twitter to implement for about 2 months now and ship out?

2 A Yes, it was.

3 Q And, at the end of the day, it took less than 5 hours to get it finalized.

4 A And shipped, yes. And it took violence occurring on the ground in order for

5 that to happen.
 129

2 [4:58 p.m.]

3 BY

4 Q And so when Twitter represents to us in this document that it began to

5 identify and review coded language that had a potential to incite violence, that reflects

6 your last minute efforts to finalize a policy that had been in the works for over 2 months?

7 A Yes. That represents -- yes. That's exactly what it represents.

8 Q And who gave you this directive to go ahead and finalize it?

9 A I was not given a directive. I was given (REDACTED) -- I was given my

10 manager's shoulder shrug as, again, a -- it felt as if the responsibility of decision drolled

11 off of their shoulder into the camera and was left on me. And that, I had to write this

12 response. Again, I was relying heavily on our legal team. And, again, I believe that Del

13 reviewed the document before it was shipped. But -- yeah.

14 Q That's where all of the back and forth in November and December about this

15 policy, at the end of the day, folks higher up in supervisory roles essentially accepted this

16 policy in a matter of hours without ever giving you a directive to formulate it or to finalize

17 it?

18 A Yes. That's exactly what happened.

19 Q And do you think that some of what happened on Twitter on January 6th

20 would have been culled more quickly or not posted in the first place had supervisors

21 implemented the policy a week in advance, 2 weeks in advance, a day in advance?

22 A Yes. I am not sure that January 6th would have been the January 6th that

23 we know it to be if the coded incitement to violence policy would have been

24 implemented in November. We would've not had, again, months of incitement to

25 violence festering on the platform and continuing to grow and to reverberate within its
 130

1 echo chamber if we would've been allowed to take that content down. It would not

2 have been sitting there on January 6th ready to go either.

3 Q I do want to get to the next entry in the timeline, but before I do, I'm

4 wondering if you can walk us through how you saw that failure to implement a coded

5 incitement to violence policy in the content on January 6th itself.

6 Were there posts you were seeing when you went in to try to track the content

7 yourself that made you regret in a micro sense -- obviously, we're having macro

8 conversations about the need for this policy. But were there instances where you were

9 like -- had targeted regret that this piece of content would not have been able to be

10 posted?

11 A Yeah. I think I felt that just kind of as an overarching feeling throughout

12 the day while balancing all of the content, right? Again, because part of what we had to

13 do was not just take down the content, the new content that was coming in on

14 January 6th; it was also taking down some of the content that'd been sitting there for

15 months. So content that I was finding and I was searching for on January 6th was not all

16 from January 6th. And so seeing that content and recognizing it didn't have to be there

17 and could have come down months ago was sobering. It was harrowing.

18 Q I understand. So we're going to scroll down a little bit to the next page.

19 We want to be mindful of the time.

20 So at the top of the page, you see there was a -- President Trump -- actually,

21 please scroll up a little bit, Heather. I want to read the tweet.

22 So at 2:24, President Trump tweeted, "Mike Pence doesn't have the courage to do

23 what should have been done to protect our country and our constitution." And he

24 derides him for not -- for refusing to halt certification.

25 And at the top of page 18, you see that Twitter labeled President Trump's tweet
 131

1 4 minutes after he posted it. And an hour after he posted it or 90 minutes

2 approximately, Twitter restricted engaging with the tweet because of claims of election

3 fraud.

4 So my first question is, how does the fact that there was only a 4-minute lapse

5 between the posting of the tweet and the labeling of the tweet cast light on how atypical

6 the typical review process of President Trump's tweets was?

7 A Yeah. So I will just say I wasn't involved in the decision or recommendation

8 around labeling this specific tweet.

9 Q Okay.

10 A So I don't have full insight into that. And I will say, 4 minutes is very, very

11 quick. And so whoever was working on that assessment was working at it at light speed

12 and at a speed that I don't know that I have seen since or before.

13 Q Okay. That is very helpful.

14 So now we can keep scrolling a little bit. You see a couple of more of President

15 Trump's tweets. And then around 3:30 p.m. Eastern, which is after President Trump had

16 issued several more tweets, Twitter says that it reinstated certain interventions that it

17 established in October 2020 that had been discontinued after the election, like

18 disengaging, disabling engagement on certain tweets.

19 What is your reaction to that action item there?

20 A It's the first I'm hearing of it. I'm -- I very much might not have been privy

21 to these discussions, and I actually don't know what interventions they are referring to.

22 Q Okay. I guess zooming back a little bit.

23 At the time, October 2020, November 2020, were you involved in any

24 conversations where individuals advocated against disabling these engagement limiting

25 policies in the first place?

 132

1 A No, I was not. I don't know what those -- I don't know what those -- what

2 the engagement disabling mechanisms are.

3 Q Okay. So going back to the time from around 2:20, 2:24 when things really

4 started to escalate. I know you talked about this earlier with Mr. Jackson, but could you

5 clarify for me again when exactly you were given the directive to, quote/unquote, stop

6 the insurrection and get the former President account suspended?

7 A Yes. That happened, I believe, around 1:30 p.m. EST.

8 Q PST?

9 A EST? EST I want to say.

10 Q Okay. Okay. So that was before he issued the tweet denouncing the Vice

11 President?

12 A I believe so.

13 Q So how did the process unfold in terms of -- we kind of talked about how the

14 coded incitement to violence policy was haphazardly up and running after months of

15 delay, but what about the framework to have President Trump removed from the

16 platform --

17 A Yeah.

18 Q -- sort of just by the mid-afternoon period?

19 A Yeah. So -- yes. So when I -- when given -- I'll be very honest with you.

20 I'm under oath as well. So when given the directive to attempt to find a reason to

21 permanently suspend the former President's account, I asked my manager why and for

22 what. And I specifically said, he's not doing anything else that he doesn't always do.

23 He's not saying anything different from what he's always said. And so I was very

24 confused as to where the directive was coming from. Again, from individuals who were

25 not allowed to touch his account for a very long time, moving from that inaction to we
 133

1 need to suspend based on the same sort of conversation and context did not make a lot

2 of sense to me.

3 My manager did not provide a lot of information and, again, was making -- was

4 not able to give me a lot of context for understanding why we should be moving in that

5 direction. However, given, again, that we were in the midst and in the thick of things, I

6 began formulating what I thought was some sort of framework that could potentially be

7 used based on what I was seeing and working on.

8 And in my mind, I started talking and sketching out the understanding of what the

9 former President was doing and saying, being outside of the public interest at this point

10 because there had been violence that was occurring. I gave -- I literally said that out

11 loud, and I gave it to the other member of the team, and they were responsible for taking

12 that argument into a variety of other rooms and making it on behalf of our team.

13 I honestly have no knowledge of any of those conversations outside of the

14 conversation I had with that original team member. That team member never gave our

15 team any sort of insight into the conversations that happened, the arguments that were

16 made, or what actually went into the determination that happened later in that day to

17 temporarily suspend the President's account and place him in a timeout for several hours.

18 Q So the last you saw of the deliberative process was the argument you

19 handed off to a team member that was basically an ad hoc rationalization about not

20 being in the public interest?

21 A Yes.

22 Q Okay. So I think at this point we'll go to exhibit -- we're on exhibit 3.

23 We'll go to page 19. Oh, actually, before we do. Sorry. Scroll back up,

24 I'm going to go off President Trump's account for a second.

25 At around 5 p.m. Eastern, so after the Capitol was -- the worst was over, Twitter
 134

1 notes that safety policy circulated additional guidance to minimize content that might

2 have the propensity to incite violence. And, essentially, it goes over some things that

3 we've talked about before, deamplifying tweets, deny listing certain hashtags.

4 Was this a policy option that had been previously considered for the 6th if things

5 had gone south?

6 A Yes. I mean, I believe what's being talked about here is the enforcement

7 guidance that I created and shipped around the world.

8 Q Oh, okay.

9 A Yeah. I think that's -- that's what I think is being talked about here. And

10 so you're asking if this was potentially a possibility to be used before this. Yes, because

11 that was almost entirely based on the coded incitement to violence policy that had

12 already been written but wasn't implemented.

13 Q So this paragraph here could have been implemented as soon as the attack

14 started, as soon as the attack was incipient, had there been a coded incitement to

15 violence policy codified at 6 a.m. Eastern on January 6th?

16 A Yes.

17 Q Okay. That's very helpful.

18 All right. So we can go back to page 19 now.

19 So right here, the second action item that is listed is President Trump's tweet,

20 which I'll read for the record.

21 At 6:01 he tweets, These are the things and events that happen when a sacred

22 landslide election victory so unceremoniously and viciously is stripped away from great

23 patriots who have been badly and unfairly treated for so long. Go home with love and

24 in peace. Remember this day forever, exclamation point.

25 What was the reaction from you and your team to this tweet when it was posted?
 135

1 A Yeah. To be really honest with you, I don't know that I really remember

2 having a specific reaction to this tweet because at that time, we were dealing with so

3 many other tweets, and we were still in the process of, I believe, bouncing tweets from all

4 around the service. And I mean, I can -- I can speak to most likely my reaction would

5 have not been great.

6 Again, the sort of go home with love and in peace and remember this day forever

7 language knowing that individuals had been shot at least at that point. I'm not sure if

8 anyone had actually died at that point, but there had been violence. And so knowing

9 that that had occurred and this -- again, the leader and main organizer of this violent

10 event is sending people home with love and peace and telling them to never forget.

11 Q So at this point, 6:01 p.m., you'd already -- Eastern time, you'd already

12 passed off the memorandum or the findings about potentially suspending President

13 Trump's account, right? It was no longer in your hands?

14 A No. That was belonging to another member on my team, yes.

15 Q Okay. So 26 minutes after this tweet, at 6:27 p.m., Twitter bounced three

16 tweets that were particularly problematic, including the one pressuring the Vice President

17 and the one right here that we talked about, and also locked President Trump's account.

18 A Uh-huh.

19 Q Did you have any part in those developments?

20 A No. Again, outside of crafting the original argument that I'm not even sure

21 if it was used within those -- in those conversations, no. I found out about this

22 happening with the rest of the world.

23 Q So you have no idea who gave the directive to do that?

24 A I do not have specific knowledge of who gave the directive to do that, but I

25 do know, if there was a directive given, it would have had to have been approved by Del
 136

1 Harvey and Vijaya Gadde, if not Jack Dorsey as well.

2 Q I guess this gets into some broader questions we talked about with the

3 unique treatment of President Trump's account.

4 If this had been someone else's account posting these things about January 6th,

5 would it have taken this long for his account to be suspended on that day?

6 A No, I don't believe so. Again, it's almost 5 hours of an insurrection

7 happening in real time before an account was locked out. In any other circumstances or

8 situations in which violence was occurring on the ground, if we knew that there would be

9 a leader or some sort of promoter of that, we would've taken it down quickly.

10 Q And do you have any idea as to why these tweets were actioned all at once

11 and his account was locked simultaneously with those actions as opposed to as they

12 occurred?

13 A I don't know why it happened as they occurred. My assumption would be

14 that, again, all of this was -- all the deliberations were happening at one time. So every

15 tweet -- they were most likely making these determinations and then adding tweets to

16 the assessment as they were coming in. And it is not -- I will say it is not uncommon for

17 Twitter to bounce several tweets together in what is called an episode.

18 And so what that means is there are several pieces of content that Twitter agrees

19 happened all within one, again, episode of time or one display. And all of those are

20 taken down at the same time together underneath one policy violation. And so this is

21 not uncommon for three tweets to come down. And as a result of that sort of

22 enforcement, the action that is taken is a subsequent -- what's called a lockout or a

23 timeout from the account. You are forced to delete the tweets that are bounced. And

24 once you delete those tweets, you have to literally sit in a timeout with a clock that

25 counts down until you're able to interact with your account again.
 137

1 Q Got it. That is helpful context.

2 And I'm wondering, in the specific framework of how Twitter dealt with President

3 Trump's account, does this activity at 6 -- at around 6:25 seem to you to be a deviation

4 from how Twitter had previously treated the President's account?

5 A Absolutely. This was the very first time that Twitter had forced Donald

6 Trump to remove his tweet. So a bounce is a forced delete, right? It locks you out of

7 your account until you delete the tweet. In every other instance in which we found

8 Donald Trump's tweets to have been violative, they were always found to, again, be

9 within the public interest, and so the public interest interstitial was applied.

10 Again, I don't know how this reasoning was reached in this circumstance and in

11 this situation, but it was absolutely outside the protocol because, again, whenever Donald

12 Trump would violate Twitter's rules, the action and enforcement that was taken was a

13 public interest interstitial, not a forced delete, not a bounce, not a timeout. This was

14 the first time this had ever occurred.

15 Q Okay. That's really helpful. And I want to pass it back to Mr. Jackson in a

16 few minutes. I have one question first that's specific and I just want to get on the

17 record.

18 When a tweet is bounced, these in particular, are they still visible on Twitter until

19 someone deletes them?

20 A I do not think so, but I think that's something to doublecheck. My

21 understanding and my recollection is that the tweet itself -- there's a tombstone that

22 pops up that says this tweet is a violation of Twitter's rules and you can no longer see the

23 tweet. But, again, it does -- Twitter as a company is not able to delete tweets

24 themselves. They have to force the user to delete the tweets themselves.

25 Q And if a user is suspended, their tweets are no longer visible?

 138

1 A If a user is permanently suspended, their tweets disappear. And after,

2 they're not -- there is not a series of record keeping that happens at Twitter to preserve

3 the record either. Once an account is permanently suspended, after a certain amount

4 of time, the data no longer exists.

5 Q Okay. At this time period, the late afternoon, early evening of January 6th,

6 were there discussions or actions taken to lock or suspend other accounts that were

7 closely associated with the President?

8 We talked earlier about sort of those influencer level people who were also

9 posting similar inciting or disinformation activity. Were you aware of any of those

10 conversations going on within Twitter?

11 A Not to my recollection. Not to my --

12 Q Do you remember --

13 A Go ahead.

14 Q Any conversation about Dan Scavino's account?

15 A Do you happen to know Dan Scavino's twitter handle?

16 Q We can probably look that up, but not off the top of my head.

17 A Okay. I'm much better with Twitter handles than real names,

18 unfortunately.

19

20 -Yes.

21 I believe it's just @DanScavino.

22 - Okay. That was going to be my guess.

23 J. Smith. Then, no. This is not ringing a bell. No.

24 BY
25 Q Okay. That's helpful. Thank you.
 139

1 So we can move away from @realDonaldTrump for one second, and go to

2 January 7th on page 19.

3 So there is a -- you can see the last action item on January 7th is that Twitter

4 allowed its coordinated harmful activities policy to permanently suspend accounts whose

5 sole purpose was to promote QAnon, and in the span of just a few days, 70,000 accounts

6 had been suspended under that policy.

7 It seems like it had -- given the number of accounts that were suspended, these

8 were not all created by January -- on January 6th or after, that would be safe to say,

9 right?

10 A Yes, that's very safe to say.

11 Q And were there detection models in place before January 6th that would

12 have allowed Twitter to do precisely this, to cull the platform of QAnon-linked accounts?

13 A Yes.

14 Q About how many months would you say Twitter would have been able to do

15 this?

16 A Let me say, so the coordinated harmful activities policy was not my policy.

17 So I will say what I know about it, and I don't have very specific and firsthand knowledge

18 of it. And that policy is essentially a -- call it like a sister policy to the terrorist

19 organization and violent extremist organizations policy. And so it is in line with some of

20 the conversation that we were having earlier around groups that may not have been

21 identified as nationalist organizations or terrorist organizations or domestic violence

22 organizations, right, but were loosely affiliated conspiracy theory networks like QAnon.

23 And QAnon had proliferated on the platform for quite some time before January 6th.

24 But it was not until the violence that happened on January 6th and its link to individuals

25 who were also linked to the QAnon community that there was a willingness to then say,
 140

1 this organization, this loosely grouped affiliation of conspiracy theorists meets the

2 threshold for what we consider to be coordinated harmful activity enough to be

3 permanently suspended.

4 Q Why do you think it took Twitter so long to take this action?

5 A It was my experience that it often took violence or death to occur in order

6 for Twitter to prioritize making policy decisions.

7 Q Okay.

8 Well, I'll pass it over to Mr. Jackson, and I think I'll have some more

9 questions after he's walked through some of the other developments in the wake of

10 January 6th.

11 Thank you, -

12

13 Q If we could return to exhibit 6, specifically page 12 of the document. This

14 looks like the right one.

15 So earlier, J., you said that you received two directives from leadership in the sort

16 of immediate aftermath of the attack on the U.S. Capitol, and those were to stop the

17 insurrection and to find the justification for the suspension of realDonaldTrump.

18 A Yes.

19 Q And Trump's account was temporarily suspended on January 6th.

20 A Yes.

21 Q Was it your understanding at that time that leadership always planned to

22 find a rationale for permanent suspension?

23 A No. That was nowhere near my belief at that time.

24 Q So you thought this temporary suspension would be lifted and he would

25 continue mostly as normal?

 141

1 A I did not believe that Donald Trump would ever be permanently suspended

2 from Twitter.

3 Q So the document we're looking at is a description of a series of meetings that

4 led to that exact outcome.

5 Did these meetings occur on January 8th or were they on the 7th?

6 A The meetings discussed in this document and all of the events that occurred

7 within this document occurred on January 8th.

8 Q Okay. What happened on the 7th to put this meeting on everyone's

9 calendar, or was there one event?

10 A Yeah. There was not an event to put this meeting or this kind of series of

11 events on anybody's calendar on January 7th. This was just kind of the aftermath of

12 Donald Trump coming out of his timeout and having to assess his tweets.

13 Q Okay. It says that your team received multiple inquiries via email and Slack

14 to conduct an assessment of the most recent tweet from @realDonaldTrump, which read,

15 The 75 million great American patriots who voted for me, America First and Make

16 America Great Again will have a giant voice long into the future. They will not be

17 disrespected or treated unfairly in any way, shape, or form, exclamation point,

18 exclamation point, exclamation point.

19 So you're called into -- after this tweet is sent, you receive sort of a barrage of

20 communications. Who was asking for this assessment?

21 A Yeah. It was a variety of different sources. I should say -- I mean, there

22 was nothing put in place, right? Like, we knew Donald Trump was coming out of the

23 timeout, but there was no monitoring in place to see what was going to be said. There

24 was no plan put in place in advance just in case there was a permanent suspension. It

25 was continuing to fly by the seat of the pants. Yes.

 142

1 Does that answer your question?

2 Q Well, it was who -- it says multiple inquires. So I'm wondering who was

3 sending the inquiries.

4 A Yes. I believe at that point we got inquiries from at least the public policy

5 team. If not, I believe we got a request and incoming from external media sources as

6 well that were interested to see what was going to occur, both internal teams as well as

7 external sources.

8 Q Okay. So Trump tweets, again, "To all of those who have asked, I will not

9 be going to the inauguration on January 20th." And your team is asked to assess this

10 tweet. Is that right?

11 A Yes.

12 Q And you found it not in violation because you did not have additional

13 context about what Trump meant by this language?

14 A Yes. Yes.

15 Q Okay. And so the document says that you were then asked by Yael Roth,

16 who is head of Site Integrity, to draft a written assessment. Would that have been a

17 normal thing to come from Mr. Roth?

18 A Absolutely not. Yael had no management power or direct supervision over

19 our team in any way, shape, or form. It was actually a very abnormal request for him to

20 be asking our team to do work for him.

21 Q So there was no -- Mr. Roth doesn't sit anywhere in the org chart between

22 you and Del Harvey?

23 A No. He reports to Del Harvey, but he, again, is on the sister team. But he

24 is not in any -- he is not even on the safety policy team.

25 Q Site Integrity is a completely different cone?

 143

1 A Yes. Different expertise.

2 Q Why do you believe Mr. Roth was the one to make this request?

3 A Mr. Roth was in a unique position at Twitter at that time. As I have

4 mentioned, there was an extreme power vacuum that had occurred after the global head

5 of safety policy's job was terminated. And the responsibility for making those decisions,

6 as we have discussed, was left up in the air for a lot of people to grab. I believe Yael,

7 knowing full well the power that our team had and the implications of these types of

8 assessments, wanted to insert himself as a way of being a part of a momentous or

9 powerful decision.

10 Q And how would you describe your position at this juncture?

11 You've just, as you said, had the responsibility for Twitter's immediate response

12 land in your lap, not something you'd asked for. You asked repeatedly for guidance

13 from a person whose responsibility it should have been. Where do you stand on

14 January 8th, 2 days later?

15 A Yes. So on January 8th, I was still the most senior and tenured person on

16 the United States safety policy team. My subject matter expertise was still in free

17 expression. And so I was the most qualified and knowledgeable individual to be making

18 these assessments on that day.

19 Q And so there was a request to write a written assessment. My

20 understanding is there were two assessments. Did you write the first assessment?

21 A I do not believe that I wrote the first assessment. I might have added some

22 language to it, but I was not the person who began drafting the first assessment, no.

23 Q And this was the assessment that found Trump to be not in violation?

24 A Yes.

25 Q And that goes to Jack Dorsey and Vijaya Gadde against -- kind of without
 144

1 your knowledge, right? This was supposed to be a draft?

2 A Yes. This was a draft that our team was working on. We had no idea that

3 it was shared all the way up with the CEO. As you mentioned, there was a first

4 assessment and another assessment. Very common within our assessments -- it was

5 very, very common for us to write multiple assessments. It was very common for us to

6 land at different decisions within those assessments. And so it was usually standard

7 procedure that we would not share an assessment unless we knew that it was finalized

8 and absolutely signed off on. And so it was a shock to us that that assessment had

9 made it all the way up the hierarchy.

10 Q So you then have a -- it looks like a 12 p.m. staff meeting with the Europe

11 team that was prescheduled --

12 A Keep scrolling.

13 Q Oh, yeah. Yeah. Please. If we could -- the next page I think.

14 A Thank you.

15 Q Sorry about that.

16 Yeah. You have a 12 p.m. assessment with the European where you discuss your

17 comfort level with that. And then later, you discuss maybe having another write-up of it

18 just in case the law enforcement team gets involved. Is that right?

19 A Yes. Yes. Again, this was pretty standard procedure at this point, right?

20 It was not uncommon especially on a decision -- this was the biggest decision, again, that

21 Twitter had ever been asked to make or be involved in. And in smaller decisions, it was

22 not uncommon for our team to reach a recommendation and then speak with our legal

23 team, and after that discussion, be recommended to either look underneath a different

24 policy or to think about additional context or something that we had not been thinking

25 about in a way that literally tipped the scale in which the assessment would go another
 145

1 direction. This was incredibly common.

2 Q And it seems like -- there are some redactions in the document. But it

3 seems like this process did actually lead to a reassessment in the work product, and you,

4 drawing on additional context, the aftermath of the political violence, the activity you're

5 still seeing on the platform, a second assessment is reached advocating permanent

6 suspension. Is that accurate?

7 A That is accurate, yes.

8 Q Okay. And then this document is sent at 1:50 p.m. for approval?

9 A Do you mind scrolling so I can see?

10 Oh, sorry. 12:50?

11 Q 1:50. Just above the meeting with Del, Yael, and some redacted names.

12 A Oh, oh, oh, oh. Yes. Okay. Yes. Yes, yes, yes.

13 Q Yeah.

14 A Yes.

15 Q And then 10 minutes later, there's a meeting with Del Harvey, Yael Roth, and

16 some other individuals.

17 And who organized that meeting? Do you have a sense of who called that

18 meeting?

19 A Yes. I believe Del called that meeting.

20 Q So at that meeting, Del questions about the team's decision -- their

21 decision making process, their position, and argues that reasonable minds could differ.

22 Could you summarize that conversation for us?

23 A Yes. I do also just want to say that it was our recommendation and we are

24 not decisionmakers. So she was talking about our recommendation, which in the

25 second assessment was for a violation of another policy. The second assessment was
 146

1 underneath a different policy and took into account more considerations, and we found a

2 recommendation to permanently suspend the former President for glorification of

3 violence.

4 You asked me about the reasonable minds -- would you mind repeating that

5 question?

6 Q Yeah. Well, if you could summarize the exchange you had with Del in

7 which she argued that reasonable minds could differ and had to be persuaded.

8 A Yes. It was very clear when I joined the conversation that Del was skeptical

9 and did not agree with the team's recommendation at that time. Del continued along

10 the lines of her magical thinking and logical leaps and, again, made this sort of statement

11 around reasonable minds being able to differ with the assessment and conclusion that

12 our team drew. Again, very much -- very skeptical and, again, very much taking a

13 continued hands-off approach to the former President and his rhetoric.

14 Q And so I have two immediate reactions to that. One, of course, reasonable

15 minds can differ, but the nature of your work was to make controversial decisions and to

16 apply judgment to cases that were very much in grey areas. Is that right?

17 A Yes. Absolutely.

18 Q So reasonable minds could differ, it would seem, on almost all of the

19 decisions that reached Del Harvey's desk?

20 A That was the gist of our job. We were the reasonable minds, and we

21 differed amongst ourselves until we came up to a recommendation that we agreed on.

22 And yes, absolutely everything that came onto our desk could go either way, which is,

23 again, why, you know, the assessment went one way and then it went the other way,

24 because it very much -- that was very much -- that was the job.

25 Q And so the document says that it was not evident at which way Del was
 147

1 leaning. Do you have any sense of what pushed this assessment over the finish line?

2 What piece of rhetoric or argument or logical claim?

3 A Yeah. I have a little bit of an idea. So the conversation that I continued to

4 have with Del was intense. And there were moments in which I reminded Del of her

5 past decisions and past precedent that she had set, and how the decision that she was

6 currently making was absolutely out of line with that past precedent, and asked her to

7 make it make sense.

8 She was not able to. This included instances in which -- Del had an ability to

9 make inferences when she wanted to and not make inferences when she wanted to.

10 Context was available to be used when she wanted to, but oftentimes context was not

11 allowed to be used. And it was -- it flip-flopped.

12 And in this specific situation, it was my argument that we could not have one of

13 those responses. And in this specific situation, one of Del's recommendations was, well,

14 why don't we just allow him to continue to tweet and we'll wait and see what happens.

15 And my response to that was, that's what we just did.

16 And I reminded her, because on January 6th, I also told Del, I am very frustrated

17 with you. I told you this was going to happen. You didn't listen to me. And so I

18 reminded her on January 8th, 2 days ago, people died, and I warned you that that was

19 going to happen. And I was looking at the service, and what I was actually seeing was in

20 response to Donald Trump's tweet about not going to the inauguration. I was seeing

21 the exact same rhetoric and the exact same language that had led up to January 6th

22 popping up underneath that tweet, but now talking about organizing and gathering for

23 another violent action to occur on January 17th.

24 And so it became very clear to me that these individuals were planning a second

25 attack. And if we allowed Donald Trump to continue to tweet in the way in which he
 148

1 had tweeted leading up to January 6th, there was going to be another attack on the

2 United States Capitol, as well as capitols around the entire United States, on

3 January 17th.

4 Del did not seem to believe me. And there was a point in which I literally said to

5 her, Del, do you want to have more blood on your hands? And at that moment, she

6 asked me to go find proof and some evidence that individuals were, in fact, planning the

7 second insurrection on January 17th. And then I left the meeting.

8 Q Was it a -- was it a tense or heated exchange?

9 A Yes. Yes. I felt like I was on trial.

10 Q Do you have a sense of how others in the room or how members of -- other

11 members of Twitter leadership at Del's level or above might have felt about this or might

12 have -- or what they might have said to her?

13 A My manager was in the room. My manager, I don't believe, was paying

14 attention to the conversation. She was, again, messaging and looking to another screen

15 on her computer. I believe that Yoel's reaction very much mirrored Del's in skepticism.

16 But, again, I did not report to Yael. I don't -- I, to this day, don't really know why I had to

17 argue to Yael or why Yael was in the room. And so I did not feel as if I needed to

18 persuade him. Del was the decision maker in the room, and it was her opinion that

19 mattered.

20 Q I have two questions I need to be sure to ask before we get off this call, and I

21 do want to make sure t h a t - h a s a chance to follow up.

22 But the first question is: If you could describe how Del's thinking and arguments

23 in this context, in this singular moment, reflect the broader pattern of Twitter's treatment

24 of Donald Trump, of Twitter's reaction to violent events, of Twitter's failure to implement

25 proactive policies, those patterns, how this moment reflects all of them or if it does.
 149

1 And then the second question is if, once the decision --

2 Ms. Ronickher. Wait. Wait. - maybe we should let her answer

3 that.

4 Sure, sure. I mean, in the interest of time, I was gonna bundle,

5 but we can just ask one.

6 J. Smith. Thank you. I have extra time if we all need it.

7 And I believe that Del's thinking, reaction, reasoning in this circumstance was

8 absolutely indicative and emblematic of Twitter's hands-off, willfully ignorant approach to

9 the former President's rhetoric on the service and on the platform. I believe it showed

10 very much, one, again, her willingness to create scenarios that did not exist in order to

11 find exemptions for plain language that was presented before you. It highlighted

12 her -- her decision making that was not consistent, and it highlighted the process around

13 that decision making that was highly inconsistent. Again, whether the ability to read

14 words into tweets occurred in one circumstance or whether it occurred in another

15 circumstance or whether it did not.

16 I also think that the wait-and-see approach is highly -- it was the Twitter approach

17 to Donald Trump, was hands off, we will wait and see what happens. And it also very

18 much shows Twitter's lack of realization and lack of wanting to reckon with the reality of

19 what happened on January 6th, as well as it's positionality, as well as its place in those

20 actions and those events that occurred.

21 It is very much a historical to say let's wait and see, when violence had just

22 occurred, and it is very much the continued approach of Twitter when it comes to these

23 circumstances as there is currently no policy, procedure, or content moderation in

24 practice that will stop this situation from occurring again.

25 BY
 150

1 Q And after the decision had been made and the ban was implemented, was

2 there ever -- before the ban was implemented or after, was there ever discussion of

3 reversing the permanent suspension?

4 A Yes and no. I say that because one of the projects that I was working on at

5 Twitter was about user rehabilitation, and the project was attempting to see how we

6 could bring back permanently suspended users onto the platform through a system or

7 process of restorative justice. It was a project that I had worked on almost my entirety

8 of time at Twitter that had gone through a variety of changes and had been essentially

9 killed at one point.

10 I want to say within a month of Donald Trump being permanently suspended from

11 the platform, we received information that there was a directive from the board of

12 directors to prioritize work about rehabilitation and restoration specifically for

13 permanently suspended users. And so while there was no explicit mention of Donald

14 Trump returning to the platform, again, the timing was incredibly coincidental. And I

15 always firmly believed that the decision could be reversed.

16 Q Was this project frog prince? That's a term that was in the notes a few

17 times.

18 A So yes. I believe project frog prince is -- was -- I believe was Trust and

19 Safety Del's response to the directive from the board in how Trust and Safety was going

20 to prioritize things like rehabilitation in order to meet those board directives.

21 Q What would a restorative justice approach to rehabilitating the former

22 President of the United States look like?

23 A It's a great question. So restorative justice is an ancient practice and

24 requires many things before participation. One includes a willingness to want to

25 participate. And a willingness to want to participate requires, one, an acknowledgment

 151

1 that harm has been done and that you have caused harm. It also includes, typically, an

2 apology and a recognition of the harm that has been done. And then it also typically

3 includes a promise of better behavior in the future [inaudible] individuals are typically not

4 even asked to participate within a restorative justice process.

5 And so for the former President, in order to be able to effectively work through

6 restorative justice process, there would need to be some sort of reflection,

7 understanding, accountability, none of which has taken place.

8 Q That's heavy. I only -- the only thing I want to do is kind of finish up the

9 chronology of events, and then I can turn it over to - f o r his followup questions

10 and some closing questions.

11 So you have this meeting with Del Harvey. You argue somewhat intensely about

12 whether or not this is something Twitter should act on immediately and whether or not

13 it's justified under the policy. And afterward, that suspension is that --your

14 recommendation is sent upstairs to Jack Dorsey and Vijaya Gadde. Is that correct?
 152

2 [5:55 p.m.]

3 The Witness. Yes. From my understanding, after we collected those tweets

4 that showed that folks were attempting and planning on gathering on January 17th, we,

5 my team and myself, worked on collecting that information, inserting it into the

6 document. And Del sent that document up her -- is that correct?

7 Hold on one second. Sorry. Let me read this.

8 Okay. I'm sorry. Before the document was sent to Vijaya and Jack, we actually

9 had another meeting. So I had another meeting with Del and Yael. The head of legal

10 was also in that meeting, and I also believe my supervisor was there.

11 And, at that point, the head of legal had actually -- am I allowed to talk about this,

12 Alexis?

13 Ms. Ronickher. No.

14 The Witness. Okay. Great.

15 The head of legal was there and added things to the document as well.

16 And, after that conversation, where it was no longer just the safety policy team

17 that was included in this recommendation but also included additional recommendations

18 from the head of legal, that document was shipped to Vijaya and Jack for their approval,

19 to my knowledge.

20 BY

21 Q Are you able to say if there was, at this stage, concern about a potential

22 lawsuit?

23 Ms. Ronickher. Only -- wait. Only answer if it was not coming from legal.

24 The Witness. From legal, I cannot speak on that.

25 I will say, however, from working at Twitter for a very, very long time, when a
 153

1 high-profile individual is permanently suspended from the service, there is a very high

2 likelihood of a suit being brought. In this specific case, it was the most famous user of

3 the service.

4 And so, in my mind, and to be honest with you, part of wanting to make sure and

5 asking that a retrospective in all of these things were put into place and making sure that

6 the information of these actions and these recommendations and determinations was

7 documented was in anticipation of a lawsuit potentially occurring.

8 BY

9 Q And, at this meeting, was the sense that this was going to happen and you

10 needed to be ironclad on process, or was this still a consultative meeting?

11 A I -- and this was, I think, around 4:00 p.m. EST. I was not 100 percent sure

12 where the decision was going. I think, given the input from legal and the extent of the

13 conversation, Del's tenor had begun to change. And so it became clear that it might be

14 a possibility, but there was no -- there was no clear directive.

15 Q And so, after this meeting, the document shipped to the chief legal officer

16 and the CEO. And, about 50 minutes later, Jack Dorsey asks Del Harvey to make edits or

17 updates to the document.

18 Since those were not fully disclosed to the team, what do you know about what

19 this document said and what the edits were?

20 A I don't know anything.

21 The document that I wrote Del copied and pasted into a new document such that

22 all of our edits and our names and histories would not appear within the document.

23 never saw that document.

24 The only information that I received, I believe, was from Del, that Jack had asked

25 for updates. But, to the extent -- I do not know -- yes.

 154

1 Q Can you summarize the document that you sent to Del?

2 A Yes. So the document that I sent to Del was a very standard, what we --

3 Ms. Ronickher. Actually, let me jump in here.

4 Was the document you sent to Del labeled "attorney-client privileged"?

5 The Witness. Yes, it was. Thank you for asking.

6 Ms. Ronickher. Now, I would say that you could answer what your

7 recommendation and the basis of it was, but not specifically what was in that document.

8 The Witness. Okay.

9 Ms. Ronickher. If that is helpful,

10 That would be fine, yeah. Thank you.

11 The Witness. Yeah. Okay. The recommendation was that the account

12 @realDonaldTrump be permanently suspended for violating Twitter's policies against

13 glorification of violence.

14 Okay.

15 Ms. Ronickher. And I just want to make clear, this was -- this safety policy --

16 The Witness. The safety policy --

17 Ms. Ronickher. Not legal's.

18 The Witness. No. It does not include legal's recommendation.

19 Okay.

20 And, then, at 6:21, you receive a message from Del Harvey that the action has

21 been taken and the account will be permanently suspended, and the decision is

22 announced publicly.

23 I have no further questions on this, but if you are willing to stick around for just a

24 little longer, I want to give Mr. Glick a chance to ask any followup and then give you a

25 chance to make any kind of concluding remarks or reflections.

 155

1 - So I think this will be very quick. I just have a couple of questions

2 about the chronology you just ran through with Mr. Jackson, and then we can move to

3 some broader conclusion questions where you can add some remarks if you wish.

4 BY

5 Q So I wanted to return to your conversation with Del Harvey about the fact

6 that the tweet in question by the former President was being used to organize what you

7 termed earlier a second attack on the Capitol on January 17th.

8 You said you searched for evidence of that attack. Can you talk us through

9 whether that evidence was difficult to find?

10 A No, it was not difficult to find. It was very easily surfaced. If you literally

11 clicked on the tweet from Donald Trump, it was in direct reply to him, again, saying things

12 like, "We're locked and loaded and ready to go, #Jan17th" or "#J17."

13 A lot of the exact same rhetoric in the public blog that was posted. I believe they

14 might have included links. I don't know if that's true. I don't know if that is -- I would

15 check that out, to see if that was included in there.

16 But, no, it was easily -- it was easily surfaced and easily found content that was

17 advocating for a second insurrection attempt on January 17th.

18 Q So is it fair to say that the former President's tweet on January 8th was being

19 used as a nexus for violent coordinated calls to action, just like his December 19th tweet?

20 A Absolutely. That is very, very fair and accurate to say.

21 Q And, once the decision came down at 6 o'clock, or a little after 6 o'clock, on

22 January 8th to permanently suspend President Trump's account, what was your reaction?

23 A I was in shock. I did not actually think that was what was going to happen.

24 By the time I had -- by the time that was announced, I had actually signed off for

25 the day and closed my computer and had essentially thought that this was going to be
 156

1 like many other decisions that had happened in which our team had made a very strong

2 recommendation and Del and the leadership team decided to go in another direction.

3 And after having, again, worked there for that long of a time and having had these

4 experiences on incredibly gray cases and incredibly huge issues, I was very prepared and

5 knew not to expect anything and, if anything, to expect a lot more of the same. And so I

6 was shocked that the action was taken.

7 Q I got it.

8 And I know you can't answer this with anything more than what you personally

9 know, but we talked a lot about the dynamics surrounding the President's account and its

10 unique treatment by Twitter. What do you think resulted in this different outcome on

11 January 8th?

12 A I think that January 8th happened and the permanent suspension happened

13 because Twitter knew -- and knew that they knew -- that they had missed the mark and

14 ignored what became January 6th.

15 And I believe that, having ignored the warnings that violence was going to occur

16 on January 6th and seeing the massive extent of violence that occurred and literally an

17 attempted insurrection occurring, I believe that that might have upped the stakes to such

18 a point that Twitter and Twitter leadership recognized that, when myself and my team

19 said again, "This is going to happen, and individuals will be harmed," we were actually

20 listened to, because I think that Twitter leadership recognized they did not want to be on

21 the hook for this happening again.

22 Q So, at the end of the day, do you think that the decision to permanently ban

23 President Trump was a result of the fact that his account was actively serving as a

24 continuing hub for coordinated calls to violence even after January 6th?

25 A Yes. Yes. I believe that the -- I believe and argued that the suspension of
 157

1 Donald Trump's account needed to occur because the language that we saw happening

2 on January 8th was a continuation and a culmination of the language that we saw

3 happening before January 6th and leading up until January 6th.

4 - Well, I think we can go towards our more general concluding

5 questions, unless Mr. Jackson or Ms. Phoenix or Ms. Connelly have any questions before

6 we go there?

7 All right. So --

8 Ms. Ronickher. Actually, I just wanted to add something with regard -- if that's

9 okay -- with regard to --

10 - Of course.

11 Ms. Ronickher. -- the attorney-client privilege. And I can hold that for the end,

12 after your questions, or I can do that now. I just didn't know general your questions

13 were going to be for concluding.

14 - Y e a h , I don't think that any of these questions are going to touch on

15 attorney-client privilege, so --

16 Ms. Ronickher. No, no. I was actually going to --

17 Oh.

18 Ms. Ronickher. Because, you know, I have instructed her not to answer

19 questions about purported attorney-client privilege, particularly with regard to, you

20 know, the memo she drafted and certain decisions. But, you know, I wanted to perhaps

21 have a couple of minutes to follow up with her, when you're all done, on the record, just

22 to clarify, you know, information about that, in case it's an issue that comes up for you all.

23 - Sure, yeah. Why don't we save that for the end, and you can --

24 Ms. Ronickher. Great.

25 Great. And thank you all for your patience.

 158

1 BY
2 Q So, given our very long conversation here today, could you summarize for us

3 your feelings about Twitter's responsibility for the violence that occurred at the Capitol

4 on January 6th?

5 A Yes.

6 I do not believe that January 6th would've happened if not for the decisions and

7 content moderation practices that were implemented, or not implemented, and put into

8 place at Twitter in relation to not just the former President but to the entire service.

9 My headphones also just died. So do you mind repeating that question while I

10 take one of them out, please?

11 Q Sure. So, given our conversation here today, could you summarize your

12 feelings about Twitter's responsibly for the violence at the Capitol on January 6th?

13 A Uh-huh. Yeah. Twitter was the home and the host for the individual who

14 was the leader and main organizer that was responsible for the attack on the Capitol on

15 January 6th. I believe that Twitter relished in their positionality of being the preferred

16 platform, by use, by the former President.

17 And their inability to disassociate and disconnect with the clout or power or

18 privilege that that esteem brought to its platform allowed it to become a cesspool of

19 incitement to violence by individuals who were not only ready but motivated and highly

20 riled on the service in order to actively commit violence.

21 January 6th was planned and orchestrated openly on twitter.com for any- and

22 everyone to see.

23 Q Given that, what actions do you think your team could have taken that

24 would've made a difference in preventing that violence and that open planning? And

25 why weren't you able to take them? Obviously, broad strokes.

 159

1 A Yeah.

2 Q We've covered a lot of this already.

3 A Yeah.

4 I believe that the safety policy team, as the team that was responsible on Twitter

5 for finding and maintaining the balance between free expression and safety, was tasked

6 with the responsibility of ensuring that the language and content that was hosted on the

7 Twitter platform did not become violative and did not spill over into a place of real-world

8 violence.

9 I believe that if our team would have been given the permissions and been

10 empowered to do our jobs the way that we knew how to do them, the way that we were

11 equipped to do them, and the way that we had always done our jobs, we would have

12 been able to create and implement a coded incitement to violence policy that most likely

13 would have removed the majority of content that was on the Twitter service that led to

14 that incitement to violence.

15 I cannot say that our team could have prevented January 6th from happening.

16 believe there are so many other circumstances and situations at play that, again, led up to

17 the full culmination of events of January 6th.

18 But I do believe that it could've happened at a smaller scale and that there could

19 have been less violence or less -- less accessibility for violence to occur if this mass

20 organizing had not been allowed to happen on the Twitter platform and service.

21 Q Thank you.

22 And so I have two more questions. And the first is, do you think that Twitter's

23 unwillingness to consider serious action, beyond interstitial messages, against President

24 Trump's account throughout 2020 contributed to the proliferation of violent calls to

25 action by the far right during the course of that year and leading up to January 6th?
 160

1 A Yes. Yes. I believe that -- sorry, my headphones just died. Can you hear

2 me?

3 Q Uh-huh.

4 A Let see if I have a headphone jack. Give me one second.

5 And we can go off the record briefly at 6:16.

6 [Discussion off the record.]

7 - And we can go back on the record at 6:16 now.

10 Q
BY-
The Witness.

Sure.
Would you remind repeating your question, please?

11 Do you believe that Twitter's unwillingness to take more dramatic action against

12 President Trump's account throughout the chaos of 2020 worsened the spread of calls to

13 coordinated violent action by the far right on the platform and on the internet more

14 broadly?

15 A Yes, very much so. I believe that -- I -- Donald Trump's rhetoric has led to

16 violence against a variety of number of people, especially on the Twitter service.

17 And I know that we are specifically talking about January 6th here, and I will also

18 say it is -- it is also -- it is large, right? It is --1 think -- my brain is thinking very much

19 about hate crimes that have happened towards members of the Asian-American

20 community specifically connected towards Donald Trump's rhetoric. And having been a

21 member of the team who argued that that content at the time should've been seen as

22 dehumanizing and as a violation of the hateful conduct policies, I see within that arena

23 how much the fanning of the flame and of the rhetoric has led to direct violence and has

24 led to direct hate crimes.

25 That sort of same approach and this Twitter's unwillingness to rein in Donald
 161

1 Trump and/or apply their standard operating procedures and policies to his account and

2 this desire, instead, to label or create interstitials I very much believe allowed the content

3 and the information to, again, proliferate. It did not stop engagement in any way,

4 shape, or form. Instead, it fanned the flames and very much allowed for these calls of

5 action to continue unfettered.

6 I believe if Twitter would have made the decision to take action and use a public

7 interest interstitial or something along the lines of -- that would have stopped

8 engagement, we would've seen a less dramatic action and it very much would have

9 helped with the spread of the content.

10 Q Thank you.

11 So my last question is forward-looking. And we've talked a lot about the ongoing

12 threat of coordinated violence being planned on social media in the immediate aftermath

13 of January 6th and in the year and a half since.

14 And my question is, as you look at the social media ecosystem today, what are you

15 most afraid of for the future?

16 A I am -- I am most afraid that there will, in fact, be a civil war that breaks out

17 within the United States and that it will occur soon.

18 I also am very much concerned, based upon my knowledge and understanding of

19 social media services and platforms, not just Twitter but the larger social media

20 ecosystem -- I am concerned, again, that January 6th will happen again. And not only

21 will it happen again, but it will happen at a bigger scale and with much more carnage than

22 what happened before.

23 Because the content moderation practices and policy philosophies that

24 undergirded the decisions that were being made at Twitter during 2020 now exist outside

25 of Twitter. They exist at small social media services, they exist at streaming services,
 162

1 they exist within gaming services. And they are within -- this understanding and

2 thinking is steeped within this ecosystem in a way in which it has proliferated and has

3 become the dominant line of thinking.

4 And without a serious reflection and understanding of how those philosophies and

5 those decisions and that approach led to direct violence and led to direct incitement,

6 those of us and those individuals who continue to work within this sector and continue to

7 work on policies and continue to work in content moderation are only doomed to repeat

8 the past, because we have not learned our mistakes.

9 The information that I am sharing with you is only known by a handful of people.

10 There is no transparency, there is no accountability within these social media services.

11 And so there is no information-sharing. There is no information- or knowledge-sharing.

12 And so each individual social media company believes itself to be different and unique

13 and special and immune from the problems and controversies of its competitors. And

14 they are not.

15 And I hope -- I hope -- that this information and being able to have the truth be on

16 the record will also allow for us to learn from the past in a way that allows us to make

17 meaningful interventions within these platforms and services in the future in such a way

18 that I hope that I am wrong, that this does not occur again.

19 And having been in a position and having said out loud, this is going to occur, I am

20 back in that position and I am telling you again: Violence and political violence will

21 occur in this country again if there are not serious changes and serious interventions

22 made within the content moderation and policy practices of social media companies.

23 Q Thank you for that. And, as with all of your answers today, that was, I

24 think, very helpful and important for us to hear.

25 So, at this point, I do want to ask if I any of my other colleagues have

 163

1 additional questions.

2 And, seeing none, I just want to ask if you, J. Smith, or, Alexis, if you'd like to say

3 anything -- Ms. Ronickher -- for the record.

4 Ms. Ronickher. Here, maybe we can go off the record just for a moment.

5 - Sure. We can go off the record at 6:25 p.m.

6 [Discussion off the record.]

7 So we'll go back on the record at 6:28.

8 And, as just mentioned, we have one more question we wanted to ask the witness

9 before we break for the day.

10 BY
11 Q And that question, J. Smith, is a followup to what you were just saying.

12 What changes do you think social media companies, big and small, need to be making in

13 order to head off the kind of conflict and violence that you are worried about?

14 A Yeah. That's a really -- that's a really, really great question.

15 I don't know if we are -- or how we honestly get to the point of making full-blown

16 change, because I very much, again, believe that we are at the point of not recognizing

17 that a problem has occurred. And I think, in order to be able to get to a place of making

18 change, we have to have that collective retrospective, to look back at the circumstance

19 and look back at the policies and procedures that were put into place or not put into

20 place before January 6th that led us here.

21 I also think that this is a tremendous case study in the rise of politics and this

22 interesting space of political influencers that has become incredibly large in social media.

23 I think that companies need to seriously understand that these individuals and that this

24 content not only exists, but the implication and the long-term impact that it can and will

25 have if left unfettered on a social media platform or service.

 164

1 I also believe that social media companies need to do a much better job of

2 speaking with each other and working in these capacities that I spoke about earlier and

3 off-service and understanding that individuals do not use the internet within a vacuum,

4 right? They exist on the internet in various corners and on various services and on

5 various platforms. And in order to get a full picture and understanding of what the

6 behavior of a user is, it is going to require a full conversation and look into the full

7 internet behavior of these individuals.

8 And I think, without that larger picture moving into these 2020 and 2024

9 elections, again, we are only bound to repeat the past, because we have not learned from

10 those mistakes.

11 Q Thank you for that answer.

12 And I think that Ms. Ronickher wanted to say something on the record

13 at this point?

14 Ms. Ronickher. Yes. I just wanted to ask a couple of questions of J. Smith with

15 regard to their answers regarding attorney-client privilege, since they did not --1 directed

16 them not to answer questions related to those.

17 J. Smith, can you describe the instructions that you were given with regard to

18 designating documents attorney-client privileged? Was it limited to communications

19 with attorneys?

20 The Witness. No. Our designation at Twitter for privileged and confidential

21 documentation was used for content that was sensitive and not wanting to be seen by

22 the eyes of the public.

23 Ms. Ronickher. And you answered a question earlier about the memorandum

24 prepared for the permanent suspension of former President Trump and to the question

25 that that was designated attorney-client privileged.

 165

1 Was that memorandum prepared for providing information to legal counsel?

2 The Witness. No, it was not.

3 Ms. Ronickher. And did legal counsel weigh in to the preparation of that

4 memorandum prior to -- in the draft that you provided to Del Harvey?

5 The Witness. I believe they may have reviewed it, but they did not provide any

6 sort of legal advice.

7 Ms. Ronickher. And so they may have reviewed it prior to the meeting with legal

8 counsel there.

9 The Witness. Yes. Yes.

10 Ms. Ronickher. Okay.

11 Those are all my questions.

12 Great. Thank you, Ms. Ronickher.

13 So, at this point, I would like to thank the witness for their time today and for their

14 willingness to work with the select committee, and would like to take a final opportunity

15 to see if anyone else has any questions on the staff?

16 Okay.

17 Seeing none, I wanted to make a final note that we will refer with counsel about

18 the issue of redactions and we will follow up on that, and the deposition will now extend

19 in recess, subject to the call of the chair.

20 And we can go off the record at 6:35 p.m.

21 [Whereupon, at 6:35 p.m., the deposition was recessed, subject to the call of the

22 chair.]
 166

1 Certificate of Deponent/Interviewee

4 I have read the foregoing _ _ pages, which contain the correct transcript of the

5 answers made by me to the questions therein recorded.

10 Witness Name

11

12

13

14 Date

15