Breaking
'MisAnthropic Statement': Who Signed It and What It Says About Anthropic
A coalition organized by the Center for AI Safety published a three-sentence statement Wednesday saying “Anthropic is not a safe AI company.” The 92 signers listed include Harvard's Steven Pinker, Hugging Face co-founder Thomas Wolf, xAI co-founder Igor Babuschkin, MIRI's Nate Soares and Extropic's Beff Jezos. It lands hours before Anthropic prices its IPO. Anthropic had not responded. Below: who signed, what each side says, and what the record shows.
The short answer
- What it is: a three-sentence statement published Nov. 18 at misanthropicstatement.com that ends: “Anthropic is not a safe AI company, and we must reject its reckless attempt to amass power in the name of the greater good.” It is an advocacy statement, not a legal or regulatory finding.
- Who organized it: the Center for AI Safety (CAIS), a San Francisco nonprofit, in a press release also distributed on Business Wire. CAIS executive director Dan Hendrycks is among the signers.
- Who signed: 92 names on the site when we checked, including Steven Pinker, Nobel economist Oliver Hart, Turing Award winner Judea Pearl, Laurence Tribe, Lawrence Lessig, Hugging Face's Thomas Wolf, xAI co-founder Igor Babuschkin, Max Tegmark, MIRI's Nate Soares and Extropic's Guillaume Verdon (Beff Jezos).
- The @MTSlive post is accurate: all four people it names (Wolf, Babuschkin, Soares, Verdon) are on the official list. Wolf, Tegmark, Pinker and Hendrycks have also said so themselves on X. We found no such posts yet from Babuschkin, Soares or Verdon.
- Anthropic's response: none found. Anthropic had not publicly responded to the statement, per MadRobot and i10X, and we found no post about it on its X account. A representative declined to comment to CNBC on a separate IPO report.
- The timing: Anthropic is expected to set its IPO price after the market closes Wednesday, with Nasdaq trading as ANT starting Thursday, Nov. 19, per CNBC. Everything in the statement is the coalition's case. Below, each claim sits next to what Anthropic's own documents say.
What we know
- The text. The preamble says “Anthropic and its rivals are racing to build superintelligence through recursive self-improvement,” and that Anthropic “uniquely portrays itself as an AI safety company and leverages that reputation to advance its commercial and political ambitions.” The statement then makes three charges: Anthropic warns about runaway AI “yet races to build it first”; it “uses safety branding to court investors, employees, and the public”; and it “has quietly weakened its safety pledges and lobbied for self-serving regulation.”
- The organizer. CAIS says it “joined” the coalition in its Nov. 18 release. Its account announced it on X at about 10:04 a.m. ET, and Hendrycks is on the signatory list. The site's contact address is a safe.ai email. A sign-up form on the site says signers are emailed to confirm their signature; we do not know what identity checks follow.
- Signers who confirmed on X. Thomas Wolf, Max Tegmark, Steven Pinker and Dan Hendrycks each posted about it on Nov. 18. Pinker wrote: “I am among the signatories of the MisAnthropic Statement.”
- Listed, no confirmation found yet. Igor Babuschkin, Nate Soares and Guillaume Verdon are named on the official list and in CAIS's release. We did not find posts from their X accounts about the statement as of mid-afternoon. That is not a denial; they may simply not have posted.
- Wolf signed with a caveat. In his post he wrote: “I don't fully agree with the wording. I wouldn't have used ‘reckless’, and I didn't know the name of the statement when I signed, which isn't my style. But I agree with the underlying sentiment.” He said Anthropic is “on track to be the most powerful company on earth” and that such power requires transparency “close to, or maybe even higher than, that of a sovereign government.” He added that he has friends at Anthropic and “a lot of respect for them in general.”
- Hendrycks's claim about silence. Hendrycks wrote on X: “Numerous AI safety groups told us privately they agreed with this statement, yet felt unable to sign without risking relationships or future funding. Anthropic has almost, but not entirely, captured the ecosystem meant to hold it accountable.” He did not name the groups. Wolf said he heard something similar from “entrepreneurs, executives and researchers.” Neither claim can be checked from outside, and Anthropic has not addressed them.
- Tegmark's reason. He wrote that he signed because he believes Anthropic is “in many ways the AI company doing the most harm.” That is his opinion. His longer post makes several further allegations that we have not verified and do not repeat here.
- The press. MadRobot, Tao Media and i10X all covered it on Nov. 18, relying on the statement site, CAIS's release and signers' posts. Each notes that Anthropic had not responded. Counts differ slightly: MadRobot said 91 signers, Tao Media 92; the list had 92 when we counted.
What we don't know
- Anthropic's reply. Whether it will respond, and whether it will do so before the listing. A company in its pre-listing quiet period may say little; we have not seen a statement either way.
- Who else signed or declined. The list is open to new signatures and was changing during the day. Whether the “numerous” unnamed groups Hendrycks mentions will sign is unknown.
- Signature checks. The site says it emails signers to confirm. How it verifies that a named person is who they say they are was not stated.
- Babuschkin, Soares and Verdon in their own words. We could not find their own comments yet, including whether they endorse every sentence as Wolf did not.
- Whether it matters to the offering. CNBC, citing unnamed sources, reported very strong demand and a possible price above the $70 to $78 range. No source we read links demand to the statement.
- Lobbying specifics. CAIS names SB 1047, New York's RAISE Act and the AEGIS Act. We could verify Anthropic's SB 1047 and federal-bill positions only through press accounts; we could not verify its positions on RAISE or AEGIS.
The statement's claims, next to Anthropic's own record
Each charge in the statement is the signers' claim. Here is what Anthropic's own pages and independent sources say about the same subject, so you can compare.
- Claim: Anthropic warns that recursively self-improving AI could “escape human control” and concentrate power. Anthropic's own essay on recursive self-improvement says: “We are not there yet, and recursive self-improvement is not inevitable. But it could come sooner than most institutions are prepared for.” It also says full recursive self-improvement “might increase the risks of humans losing control over AI systems.” We did not find “concentrate power” language on that page.
- Claim: yet it “races to build it first.” In its IPO prospectus (SEC filing, amended Nov. 17), Anthropic describes each Claude model helping build the next, which it calls “closing the loop,” and says “we believe it is one of the most significant structural advantages in frontier intelligence.” The filing's risk factors also say its position at the frontier “depends in part on” this AI-assisted research. Anthropic's essay says AI is already accelerating AI development, while full autonomous successor-building has not been reached. Whether that amounts to “racing to build it first” is the signers' characterization.
- Claim: it “uses safety branding to court investors, employees, and the public.” Anthropic's X profile calls it “an AI safety and research company.” The prospectus says it is a public benefit corporation “consistent with our long-term commitment to AI safety and responsibility,” with a Long-Term Benefit Trust that can elect and remove a majority of the board. Whether that is substance or branding is the central dispute; the filing documents the structure, not the motive.
- Claim: it “quietly weakened its safety pledges.” CAIS says that in February Anthropic weakened its Responsible Scaling Policy, “including fully abandoning its unilateral commitments to pause.” Anthropic published the new policy openly on Feb. 24, 2026 in its RSP v3.0 post. That post says some safeguards at higher capability levels “might prove outright impossible to implement without collective action,” and that the policy now aims for “more realistic unilateral commitments.” The post does not use the word “pause.” The independent GovAI analysis says: “Anthropic also dropped its pause commitment. But importantly, Anthropic is not lowering any of its existing mitigations.” GovAI was initially negative, then said that on balance it is “better to be honest about constraints than to keep commitments that won't be followed in practice.” So the pause commitment was dropped, publicly; whether that is “quiet” or a weakening is disputed.
- Claim: it “lobbied for self-serving regulation.” CAIS says Anthropic lobbied to weaken SB 1047, the RAISE Act and the AEGIS Act and privately advocated federal preemption of state AI laws. On SB 1047, Business Insider reported in 2024 that Anthropic proposed amendments in July, then wrote to Gov. Gavin Newsom that the bill's “benefits likely outweigh the costs,” adding “we are not certain of this.” On federal preemption, Tao Media reports, citing Politico and a law-firm analysis, that a draft federal bill Anthropic wrote would override some state rules, and that Anthropic's federal affairs lead said the language was “not precise enough” and would be narrowed (we could not read the Politico story, which blocked our tools). Anthropic's own RSP post, by contrast, cites SB 53 and the RAISE Act as examples of laws that “start to require” safety frameworks, which it says is what the RSP set out to encourage.
The IPO context
- The offering. The prospectus, dated Nov. 17, covers 825,000,000 Class A shares with an indicated price range of $70 to $78, and says Anthropic has been approved to list on the Nasdaq Global Select Market as “ANT.” At the range's ends that is roughly $57.8 billion to $64.4 billion, matching CNBC's figure of up to $64.35 billion at the high end.
- Latest reporting. CNBC reported Wednesday morning, citing people familiar, that Anthropic is considering upsizing the deal and pricing “well above $80,” with institutional demand around $400 billion and retail around $100 billion. It noted the orders are indications of interest, not binding, and that the final price comes after the close Wednesday, with trading Thursday.
- Hendrycks's framing. CAIS's executive director said: “You cannot credibly claim to be safe while weakening your own previous safety commitments and asking the public for tens of billions of dollars to build recursively self-improving AI.” That is the coalition's argument, timed to the offering.
Who signed, and why their motives differ
The list includes direct competitors, an AI-risk research group's president, accelerationists and academics. Here is how the groups line up, with the source for each. This is context for reading the list, not a verdict on any signer.
- AI-risk and pause advocates. Nate Soares is president of the Machine Intelligence Research Institute (MIRI), and signed as an individual; his X bio reads “Trying to make AI not kill everyone.” Max Tegmark and Anthony Aguirre of the Future of Life Institute, and leaders of PauseAI and ControlAI, are also listed. Many in this group want frontier AI development slowed or paused. MadRobot reports that one signer's group, Irreplaceable, is among four nonprofits suing Anthropic and four other labs to freeze frontier AI.
- Competitor-linked signers. Igor Babuschkin is listed as co-founder and CEO of River AI and co-founder of xAI, which MadRobot calls a direct Anthropic rival. His X bio says he was previously at xAI. Dan Hendrycks's X bio lists past advisory roles at xAI, Scale and Meta AI; he runs CAIS, the organizer.
- Accelerationists. MadRobot describes Guillaume Verdon (Beff Jezos), the Extropic founder, and George Hotz of the tiny corp as “accelerationists who want AI built faster.” CAIS calls the coalition “AI safety researchers, open-source advocates, AI ethicists, and accelerationists.”
- Open-source, ethics, law and science. Hugging Face co-founder Thomas Wolf, Rumman Chowdhury of Humane Intelligence, law professors Tribe and Lessig, the Authors Guild's Mary Rasenberger, Nobel laureates Oliver Hart and Sheldon Glashow, and several Fields medalists are among the rest. The list gives titles, not reasons.
- Why it matters for reading the list. Signers share three sentences, not a program. Some think AI is dangerously fast, some think Anthropic's safety politics are shielding it from competition, and one has said he would not have used the word “reckless.” Their different motives do not make the claims true or false; each claim is checked above against the record.
The full list
The official site lists every signer with a title and affiliation, and accepts new signatures: misanthropicstatement.com. CAIS's release names a shorter set of prominent signers. Titles on the list are as given by the site and may reflect each signer's role, not their employer's view.
Get the daily issue
One email a day. Unsubscribe anytime.
An Underclassic breaking story, updated as reporting develops. Daily briefing: underclassic.com. Related: our breaking story on the Trump administration's Chinese-AI disclosure idea; AI API pricing per 1M tokens. Sources: The MisAnthropic Statement; CAIS press release, Nov. 18, 2026 (Business Wire); X posts by CAIS, Dan Hendrycks, Thomas Wolf, Max Tegmark, Steven Pinker and @MTSlive; MadRobot; Tao Media; i10X; CNBC, Nov. 18, 2026; Anthropic prospectus (SEC); Anthropic RSP v3.0; Anthropic on recursive self-improvement; GovAI on RSP v3.0; Business Insider, Aug. 2024. Signer counts and X posts were checked on Nov. 18, 2026. Every allegation above is attributed to the person or group making it.