18 November 2026 · Center for AI Safety / signatories
The MisAnthropic Statement: 'Anthropic is not a safe AI company', signed on IPO day
On 18 November a coalition around the Center for AI Safety published the MisAnthropic Statement, three sentences saying that Anthropic is not a safe AI company. The page listed 89 names that afternoon and 116 by 04:11 UTC on 19 November, among them Nobel laureates, the Turing laureate Judea Pearl, and co-founders of Hugging Face and xAI. The statement cites no sources, and no response from Anthropic turned up in the coverage read for this entry.
By Mara Masaeva · maramasaeva.comUpdated 19 November 2026
ArgumentDevelopingThe story is still developing. There is one source so far, or the numbers are still changing.
What happened
The text says three things. Anthropic warns that AI which improves itself could concentrate power and escape human control, yet races to build it first. It uses safety branding to win over investors, employees and the public, while having weakened its safety pledges and lobbied for regulation that suits itself. And it concludes that Anthropic is not a safe AI company, and that its attempt to amass power in the name of the greater good should be rejected. The page names no organiser; the contact addresses are at safe.ai, the domain of the Center for AI Safety, and CAIS posted the call to sign on X.
The signers disagree about a lot. The page lists the psychologist Steven Pinker, the law professors Laurence Tribe and Lawrence Lessig, the Nobel laureates Oliver Hart and Sheldon Glashow, Judea Pearl, Max Tegmark, Douglas Hofstadter, Gary Marcus, Roman Yampolskiy and CAIS director Dan Hendrycks. Also on it are Hugging Face co-founder Thomas Wolf, xAI co-founder Igor Babuschkin, and Guillaume Verdon, a founder of the accelerationist e/acc movement. The National adds Tristan Harris of the Centre for Humane Technology, and Digg adds Rumman Chowdhury.
The count moves, and outlets counted at different times. The National wrote 'at least 80', two outlets counted 89 on the page around 16:10 UTC on 18 November, another counted 92, and SnowRock saw 116 on the morning of the 19th. Do not average these: the list was growing. Times of AI puts it at 92 by early on 19 November, which sits oddly beside SnowRock's 116 at 04:11 UTC; the sources do not explain the gap.
Two signers explained why they signed. Max Tegmark says Anthropic's leaders warn in public about recursive self-improvement, the process of AI building better AI, while in private they push it and call it a competitive advantage to investors. His list of ways Anthropic does more harm than competitors also says Claude helped select a school as a target in Iran through the Palantir collaboration. Dan Hendrycks says AI safety groups told CAIS privately they agreed, but did not dare sign without risking relationships or future funding, and that Anthropic has almost, but not entirely, captured the ecosystem meant to hold it accountable. Both are claims by people who signed.
Anthropic did not respond when The National asked on 18 November, and SnowRock found no public response by 04:11 UTC on the 19th. The Next Web and BigGo also report no public response by Thursday morning. On listing day Daniela Amodei told CNBC that Anthropic has been a safety-first company from day one; the pages read do not connect that remark to the statement.
What came before
The signers' premise is the February rewrite of Anthropic's Responsible Scaling Policy, in which the pledge not to train a model without guaranteed safeguards was dropped, and the listing it was timed against.
What may follow
A reading, not a fact: the statement does not say which lobbying it means. The draft bill Anthropic wrote itself in October is one candidate, but the page does not make that link.
What I do not know
The statement is an opinion signed by many people, not an investigation. It cites no sources, has no FAQ, and does not say which pledges it means. That is why this entry is marked 'check'.
It says Anthropic 'quietly' weakened its pledges. The record in the next entry is that the change was announced in a blog post and covered by Time, Bloomberg and others, so 'quietly' is arguable.
Tegmark's claim about the school is contested. In the Iran campaign, Maven, Palantir's targeting system, ran on Claude, according to the Washington Post, which has not been read here. A US missile strike on the school at Minab on 28 February killed between 150 and 175 people, depending on the source. Semafor reported on 18 March that humans, not AI, were to blame, through errors in the target lists, and Dario Amodei told Bloomberg he does not know what role Claude played. Which part Claude played, if any, is not established.
Individual names on the list do not mean their organisations endorse it, as Digg also notes.
Read next
- Anthropic prices its IPO at $83 and lists on Nasdaq today: $81.3 billion, the biggest ever
- Anthropic drops its pledge to stop training if safeguards fall short: Responsible Scaling Policy v3
- Rally outside Anthropic on listing day: 'Ban RSI', a ban on AI building better AI
- A year of talking to Anthropic employees: Jasmine Sun in the New York Times
- Anthropic writes its own AI bill: a stop only above 50 victims or a billion dollars
Sources
- The MisAnthropic Statementprimary · main source
The statement and its signer list. Source for the text and the names listed. Gives no sources of its own and no organiser, only safe.ai contact addresses.
- The National: Anthropic IPO not worth 'risking lives of loved ones', petitioners saypress
18 November 2026. Source for 'at least 80' signers, Pearl, Tegmark and Harris, Tegmark's argument, and that Anthropic did not respond.
- SnowRock: Anthropic IPO prices at $83 into a safety revoltpress
Source for 116 signers at 04:11 UTC on 19 November, CAIS announcing the statement, and that no Anthropic response was found.
- FourWeekMBA: 89 sign MisAnthropic Statement hours before Anthropic IPOpress
Source for the count of 89 at about 16:10 UTC on 18 November.
- Cellcog: MisAnthropic Statement, 92 sign against Anthropicpress
Source for the count of 92. One of several counts taken at different times.
- Digg: Hendrycks says AI safety groups backed MisAnthropic in private, but feared signingpress
Aggregator. Source for Hendrycks's claim about funding fears and for Chowdhury on the list. Hendrycks's claim is his own and unconfirmed.
- The Next Web: Nobel and Turing winners sign statement calling Anthropic unsafe before IPOpress
Source for CAIS organising the statement and for no public Anthropic response by Thursday morning.
- BigGo Finance: Nobel laureates slam Anthropic as 'not a safe AI company' on IPO daypress
Second report that Anthropic had issued no public response.
- Times of AI: Anthropic faces criticism from 89 signatories ahead of IPOpress
Source for 89 at first circulation and 92 by early 19 November.
- Mayhem: What are you buying with the Anthropic IPO?argument
Opinion newsletter. Used only for Verdon and Lessig as signers.
- Semafor: Humans, not AI, are to blame for deadly Iran school strike, sources saypress
18 March 2026. Counterweight to the claim about Claude and the school. Based on anonymous sources.
- Let's Data Science: Anthropic CEO addresses possible Claude role in Iran school strikepress
Source for Amodei telling Bloomberg he does not know what role Claude played. Secondary report of the interview.
- Washington Post: Anthropic AI in the Iran campaignpress · not read end to end yet
4 March 2026. Cited by others for Maven running on Claude. Linked, but I have not read it.