Article
The Twitter Files (December 2022 onward) and Murthy v. Missouri, 603 U.S. 43 (2024)
The Twitter Files are internal Twitter emails and messages that the company's owner gave to selected writers, beginning with Matt Taibbi on 2 December 2022, who published them as threads. The audit covers the threads on the Hunter Biden laptop story and on FBI contacts (parts 1 and 7) and Taibbi's House testimony. It sets them beside the record those contacts produced in Missouri v. Biden: the W.D. La. ruling of 4 July 2023, the Fifth Circuit opinion of 3 October 2023, the Chan deposition, the Supreme Court's decision of 26 June 2024 resting on standing, and the March 2026 consent decree. The threads are screenshots with commentary; the court record is sworn testimony and judicial findings.
Who made it
- The Twitter Files documents were given to selected writers by Twitter after Elon Musk acquired the company; Taibbi's written testimony says the first contact was a note from an online source, and the first thread appeared a week later (Taibbi testimony, 9 Mar 2023, p. 1).
- Matt Taibbi is the editor of Racket on Substack and a former Rolling Stone writer (Taibbi testimony, p. 1). Michael Shellenberger and Bari Weiss are independent writers who joined the releases a week after the first thread (same, p. 1).
- Taibbi's written testimony says the review covered 'tens of thousands of emails' (p. 2). No complete archive of the documents has been published; the threads present selected screenshots.
- Murthy v. Missouri was brought by the States of Missouri and Louisiana and five individuals (Bhattacharya, Kulldorff, Kheriaty, Hines, Hoft) against dozens of federal officials and agencies. Bhattacharya and Kulldorff later withdrew as plaintiffs on taking federal posts (Consent Decree, Doc. 478, n. 1).
- Decision of the Supreme Court: 6 to 3, Barrett for the Court, Alito dissenting with Thomas and Gorsuch (Murthy, syllabus p. 5).
Its claims checked against the record
10 claims checked:3 supported · 6 partly supported · 1 contradicted · 0 not in the record
A verdict rests on the documents named in its row.
| The work says | Verdict | What the record shows |
|---|---|---|
| Twitter blocked links to the New York Post's Hunter Biden laptop story on 14 October 2020 under its hacked-materials policy. The files show Twitter staff doubting that policy basis and show no government involvement in that decision, only a general earlier warning from federal law enforcement about foreign hacks. | Supported by the record | The District Court records that Twitter blocked users from sharing links to the story and barred users who had shared it from tweeting until they deleted (Memorandum Ruling pp. 63-64). The Supreme Court states that 'multiple platforms suppressed a report about Hunter Biden's laptop, believing that the story originated from a Russian hack-and-leak operation' (slip op. p. 3). Part 1, as read, states that no evidence of government involvement in the decision was seen, and Part 7 reproduces Roth writing on 14 October that the material did not clearly violate the policy but 'feels a lot like a somewhat subtle leak operation'. |
| Before the 2020 election the FBI and other agencies repeatedly told Twitter and the other platforms to expect a state-actor hack-and-leak operation, met them regularly, sent material through a one-way FBI channel called Teleporter, and Twitter staff came to expect such an operation. | Supported by the record | The Supreme Court states: 'Shortly before the 2020 election, the FBI warned the platforms about the potential for a Russian hack-and-leak operation. Some companies then updated their moderation policies to prohibit users from posting hacked materials' (slip op. p. 5). Roth's sworn declaration of 17 December 2020 says federal agencies 'expected hack-and-leak operations by state actors might occur' shortly before the election, 'likely in October' (Memorandum Ruling p. 62). Chan testified the FBI used Signal and Teleporter and that it repeatedly asked platforms whether they had changed their hacked-materials policies, while denying it urged changes (pp. 61-62). |
| FBI officials primed Twitter to treat a Hunter Biden leak as a Russian hack-and-leak, although the FBI had held the laptop since December 2019. | Partly supported | Roth's declaration says he 'learned in these meetings that there were rumors that a hack-and-leak operation would involve Hunter Biden' (Chan dep. p. 217; Memorandum Ruling p. 62). Chan testified that Hunter Biden was not referred to in the CISA industry meetings, that he does not remember saying the name to Twitter, and that Hunter Biden came up once, at a Facebook meeting after the story broke, where Dehmlow said the FBI had no comment because it had not confirmed an investigation (pp. 213-216, 228-229). Asked whether Dehmlow knew the laptop contents had not been hacked, he answered 'I have no idea' (p. 215). The District Court found the FBI had the laptop from 9 December 2019 and 'likely misled' the companies (Memorandum Ruling pp. 63, 107-108); the Fifth Circuit's FBI holding rests on takedown requests and policy changes and does not mention the laptop (pp. 54-57). |
| Twitter collected $3,415,323 from the FBI between October 2019 and early 2021 under a reimbursement program, which the coverage of part 7 presented as payment tied to the FBI's moderation requests. | Partly supported | The figure comes from a redacted 10 February 2021 email from a Twitter employee to Twitter's then-deputy general counsel and general counsel that describes 'a reimbursement program for our legal process response from the FBI' begun in 2019. The statute requires agencies to reimburse 'such costs as are reasonably necessary' for responding to Stored Communications Act requests (18 U.S.C. 2706(a), quoted in the D.D.C. opinion p. 2). The court describes the invoices as created 'so that the FBI could pay for warrants, subpoenas, and other legal-process requests' (p. 6). No document read ties the payments to flagged-content requests. |
| Twitter, Facebook, Google and others took moderation requests from every corner of government: the FBI, DHS, HHS, DOD, the State Department's Global Engagement Center, and the CIA. | Partly supported | The Fifth Circuit found regular contact and content flagging by the White House, the Surgeon General's office, the CDC, the FBI and CISA (pp. 43-60). It found NIAID officials had 'insignificant (if any)' communication with platforms and that State Department communications 'educat[ed] the platforms' about foreign actors' tools, with 'no indication' that State officials flagged specific content (pp. 60-61). Neither the District Court ruling, the Fifth Circuit opinion nor the Supreme Court opinion mentions the CIA or the Department of Defense as sources of requests; the claim about those two agencies is outside the record read. |
| The FBI sent Twitter and other platforms lists of accounts and URLs to review, and the platforms acted on them. | Supported by the record | Chan testified that after the FBI shared information, platforms would report back that they had taken accounts down: 'I would say it was like a 50 percent success rate. But that's just from my recollection', counting any action taken as a terms-of-service violation (dep. p. 167). The District Court records that the FBI made no attempt to distinguish foreign from American sources in these reports and that it sent indicators by Teleporter one to five times a month (Memorandum Ruling p. 65). The FBI says it 'never direct[s] or ask[s]' platforms to act (FBI statement to FactCheck.org); the Fifth Circuit held the requests were 'reasonably construed' as coercive because of the FBI's authority (pp. 55-56). |
| A federal court found that the government coerced or significantly encouraged the platforms' moderation of speech. | Partly supported | The District Court found plaintiffs 'likely to succeed' and described 'a far-reaching and widespread censorship campaign' (p. 154). The Fifth Circuit held the White House and Surgeon General's office, and the FBI, likely coerced and significantly encouraged moderation, and that the CDC and CISA likely significantly encouraged it, but removed NIAID and the State Department from the injunction (pp. 43-61). These are preliminary-injunction findings on likelihood. The Supreme Court reversed for lack of standing, said the Fifth Circuit 'glossed over complexities in the evidence', and expressed no view on the merits (syllabus p. 3; n. 3). The case ended by consent decree, which states it is not 'evidence or... an admission regarding any issues of law or fact' (para. 17). Record not produced. Discovery to establish standing, permitted by order of 8 Nov 2024 (ECF 404), was stayed on 3 Feb 2025 (ECF 432), and the case ended by consent decree on 25 Mar 2026 with no merits findings and no admission, so no evidence beyond the preliminary-injunction discovery was produced (Consent Decree, Doc. 478, paras. 5, 10, 17). |
| The Supreme Court in Murthy v. Missouri decided the censorship question, either that the government did not censor or that it did and the suit failed anyway. | Contradicted by the record | The Court held 6 to 3 that 'neither the individual nor the state plaintiffs have established Article III standing to seek an injunction against any defendant' (syllabus p. 2). Footnote 3: 'Because we do not reach the merits, we express no view as to whether the Fifth Circuit correctly articulated the standard for when the Government transforms private conduct into state action' (slip op. p. 7). The consent decree repeats that the majority 'did not reach the merits' (para. 3). |
| Each plaintiff's platform restrictions were traceable to demands by the government defendants. | Partly supported | The majority found the record lacked 'specific causation findings with respect to any discrete instance of content moderation', that Facebook began restricting Hines's COVID content before most of its communications with the White House and CDC, and that Hoft's own declaration shows Twitter acted under its rule against sharing private intimate media, on a post by his brother Joe Hoft (pp. 10-18). It called Hines's showing the best of the plaintiffs and still 'tenuous' (syllabus p. 3). Three justices dissented, would have found Hines 'indisputably injured' by coerced Facebook enforcement, and said Facebook 'repeatedly yielded' (Alito, J., dissenting). The majority says the dissent draws links 'Hines herself has not set forth' (n. 7). |
| The platforms acted under government pressure, not their own judgment, and have since said so. | Partly supported | Zuckerberg wrote that in 2021 senior Biden Administration officials 'repeatedly pressured our team for months to censor certain COVID-19 content, including humor and satire' and that 'the government pressure was wrong'; he also wrote 'Ultimately, it was our decision whether or not to take content down', and that Meta 'temporarily demoted' the Post story while fact-checkers checked it (Mace letter p. 1; Fortune; AFP). Alphabet's counsel wrote that White House officials 'conducted repeated and sustained outreach' and 'pressed the Company regarding certain user-generated content' that 'did not violate its policies' (Gizmodo, quoting the letter). The Supreme Court found the platforms had independent incentives and began suppressing plaintiffs' COVID content before the challenged communications (syllabus p. 3). |
What the record shows that it leaves out
- The platforms' own policies pre-date the contacts: Facebook began fact-checking election claims in 2016 and removing health misinformation in 2018, YouTube announced demonetization of anti-vaccine channels in 2019, and Facebook, Twitter and YouTube applied COVID-19 policies from January, March and May 2020 (Murthy, slip op. p. 2).
- The Supreme Court found that the 'frequent, intense communications' of 2021 between the White House and the platforms 'had considerably subsided by 2022' and that CISA's switchboarding ran 'until mid-2022' (Murthy, syllabus p. 4; slip op. p. 5); the Fifth Circuit wrote in October 2023 that the contact 'continues to this day' (83 F.4th at 2).
- Chan's account that Hunter Biden came up in a single FBI-Facebook meeting after the story broke, and that Dehmlow's 'no comment' rested on the FBI not having confirmed an investigation (Chan dep. Doc. 204-1, pp. 213-216).
- Taibbi's part 1 states that requests to remove content came from the Trump White House as well as the Biden campaign and were honored by Twitter (as read in the Thread Reader copy); the court record concerns only the Biden administration's contacts, and the House hearing record held here has no comparable material on 2017-2020 contacts.
- Twitter's own 2021 audit of algorithmic amplification found that in six of seven countries the mainstream right was amplified more than the mainstream left (Huszar et al., PNAS, exhibit Exhibit 722). It measures amplification, not removals, and bears on the District Court's description of the alleged censorship as 'almost exclusively' of conservative speech (Memorandum Ruling, introduction).
- The Fifth Circuit's finding that the CDC did not coerce but significantly encouraged the platforms by defining what counted as misinformation in 'Be On the Lookout' meetings and answering platform questions about specific claims (ca5 pp. 57-59); this is the strongest documented instance of the agency-defines-truth pattern and appears in no Twitter Files thread read.
- The FBI's quarterly payments to Twitter for 2016 to 2023 remain redacted under FOIA Exemption 7(E) (Judicial Watch v. DOJ, D.D.C. 4 Feb 2026, Doc. 27, pp. 1-2); the Twitter Files figure covers only October 2019 to early 2021.
- The consent decree of 25 March 2026 binds only the Surgeon General, CDC and CISA, for ten years, only as to the plaintiffs' own accounts and five named platforms; it permits the government to give platforms information 'the companies are free to use as they wish' and to say that posts are inaccurate (Consent Decree paras. 23-26).
- Executive Order 14149 (20 Jan 2025) directed the Attorney General to investigate the prior four years' activity and report with recommendations; the consent decree recites it (Consent Decree paras. 6-8). No report from that directive was found.
Documented efforts and withheld records
Each item is a documented act, with its source: an act against the work, its author or its sources, or the withholding of records on its subject. The list states what was done and does not state why.
- 14 Oct 2020. Twitter blocked users from sharing links to the New York Post's Hunter Biden laptop story and prevented users who had already shared it from tweeting until they deleted the tweets; Facebook 'temporarily demoted' the story pending fact-checking. Missouri v. Biden Memorandum Ruling, W.D. La., Doc. 293, pp. 63-64; Zuckerberg letter to Rep. Jordan, 26 Aug 2024 (as reported by Fortune)
- Dec 2020. Twitter suspended the account of Joe Hoft, brother of plaintiff Jim Hoft, after a post of Hunter Biden photographs, citing its rule against sharing private intimate media without consent (the stated rule comes from Hoft's own declaration as quoted by the Court). Murthy v. Missouri, slip op. pp. 16-17
- Oct 2020 to Apr 2023. Facebook fact-checked plaintiff Jill Hines's posts from October 2020, restricted her account in May 2022 and issued a warning in April 2023; the Court found only some of these linked to government communications. Murthy v. Missouri, slip op. pp. 17-21
- 9 Mar 2023; letters 27 Mar and 24 May 2023. An IRS revenue officer visited Taibbi's New Jersey home on the day of his House testimony and left a note. The IRS told him the reason was identity-theft rejections of his 2018 and 2021 electronic filings; Taibbi said neither he nor his accountant received the 2019 and 2020 letters the IRS cited. The Judiciary Committee's letter states the IRS production shows the 2018 examination opened on 24 Dec 2022 and that the IRS did not produce the cited letters. The Committee's letters stated the visit 'could be interpreted' as intimidation; the IRS stated its identity-theft reason. Letters, Rep. Jordan to Treasury Secretary and IRS Commissioner, 27 Mar 2023 and 24 May 2023
- 7-8 Apr 2023. Twitter, under Musk, throttled or labeled links to Substack, Taibbi's publisher; Substack's chief executive said links were severely throttled and Taibbi announced he was leaving the platform. Musk replied that Substack links were never blocked and that Taibbi's statement was false. Racket statement by Taibbi; Mediaite and Fox News reports of Musk's reply
- 12 Apr 2023 subpoena; letter 24 Jun 2024. The House Judiciary Committee subpoenaed the Stanford Internet Observatory, which ran the Election Integrity Partnership named in the court record, and stated that Stanford did not produce the Jira ticket data by the 28 Apr 2023 return date. The Committee's June 2024 letter states that, according to public reporting, the Observatory would be discontinued. Letter, Rep. Jordan to counsel for Stanford University, 24 Jun 2024
- Request 2022-2023; ruling 4 Feb 2026. The FBI refused to produce or confirm records of payments to Twitter in response to a Judicial Watch FOIA request, produced 44 redacted pages after suit, and the D.D.C. upheld withholding of the quarterly payment totals for 2016 to 2023 under Exemption 7(E). Judicial Watch v. DOJ, No. 23-3004 (JEB), Doc. 27, pp. 1-2
- 8 Nov 2024 to 25 Mar 2026. Discovery in Missouri v. Biden ended without merits findings: standing discovery allowed on 8 Nov 2024 was stayed on 3 Feb 2025, and the court entered a consent decree on 25 Mar 2026 releasing all claims and stating it is not evidence or an admission. Consent Decree, Missouri v. Biden, Doc. 478, paras. 5, 17, 31
Published responses
- FBI spokesperson: Said the correspondence with Twitter shows 'nothing more than examples of our traditional, longstanding and ongoing federal government and private sector engagements', that the FBI provides information so companies can protect themselves, and that 'conspiracy theorists and others are feeding the American public misinformation'. FBI told FactCheck.org it 'never direct[s] or ask[s]' companies to take action.
- Federal defendants in Missouri v. Biden: Argued in the Fifth Circuit that officials 'sought to mitigate the hazards of online misinformation' by 'calling attention to content' that violated platform policies, a form of government speech. In the consent decree the government agreed to a ten-year injunction on three agencies without admission of any fact or law (para. 17).
- House Judiciary Committee Democrats, 9 March 2023 hearing: Members called the witnesses 'so-called journalists' and said the Twitter Files had produced no genuine example of the government censoring lawful speech (reported).
- Meta (Mark Zuckerberg letter to Rep. Jordan, 26 Aug 2024): Wrote that the government pressure was wrong, that Meta regrets it was not more outspoken, and that it was ultimately Meta's decision whether to take content down. White House spokesman: 'Our position has been clear and consistent' (reported by Fortune, citing Politico).
- Alphabet (letter to Rep. Jordan, 23 Sep 2025): Wrote that White House officials 'conducted repeated and sustained outreach' about COVID-19 content that 'did not violate its policies', and said YouTube would give creators terminated under policies 'no longer in effect' an opportunity to rejoin.
- Kate Starbird, University of Washington Center for an Informed Public: Replied to the House Judiciary report on CISA that her CISA advisory subcommittee had no contact with CISA staff in 2020, did not alert platforms to content, and that the report conflated timelines and quoted emails selectively.
- Twitter (Dorsey and Gadde, October 2020): Reported at the time: Twitter said its communication about the Post story was not good and that blocking URLs without context was wrong, then changed the hacked-materials policy to label instead of block; the Post article stayed blocked under a separate private-information rule. Read only through news summaries.
Where it connects
- Managing the Narrative: Documents agency contacts with platforms about content: the CDC misinformation advisories, FBI account reports and CISA switchboarding, and the courts' and platforms' stated positions on them.
- Foreign Funding and Influence: The FBI Foreign Influence Task Force and CISA election-security meetings were the channel for the 2020 hack-and-leak warnings, and the State Department's Global Engagement Center is named in the record.
- Division as Revenue: Twitter's own amplification audit and the platforms' pre-existing moderation policies bear on what the platforms did for their own reasons and on partisan asymmetries in the record.