A coalition of scientists, legal scholars, and tech figures has called Anthropic “not a safe AI company”. They did it a day before the Claude maker’s shares start trading. The MisAnthropic Statement went up on Wednesday, organised by the Center for AI Safety (CAIS), a San Francisco nonprofit.
The signatories rarely agree on anything about AI. They include AI safety researchers, open-source advocates, and AI ethicists. Accelerationists, who want AI built as fast as possible, signed too.
“Anthropic warns that recursively self-improving AI could concentrate power and escape human control, yet races to build it first,” the statement says.
Recursive self-improvement means AI systems doing the work of building better AI systems. The statement accuses Anthropic of using its safety branding to court investors, staff, and the public. It ends by urging people to reject what it calls Anthropic’s reckless attempt to amass power.
The names on the site include Harvard psychologist Steven Pinker, Nobel economist Oliver Hart, and Turing Award winner Judea Pearl. Harvard law professors Lawrence Lessig and Laurence Tribe are on it too. Fields medallists Peter Scholze, James Maynard, and David Mumford signed too.
From tech, Hugging Face co-founder Thomas Wolf, xAI co-founder Igor Babuschkin, and George Hotz signed. So did Max Tegmark of the Future of Life Institute. His name sits next to Guillaume Verdon, who founded the e/acc movement as “Beff Jezos”.
“People in this coalition disagree on almost everything, yet we all recognize the danger of a company that convinces itself its own dominance is the greater good,” Dan Hendrycks, executive director of CAIS, said in a statement.
In its announcement , CAIS set out three charges. It says Anthropic warns about recursive self-improvement while pitching it to public-market investors as a core advantage.
It also says Anthropic weakened its Responsible Scaling Policy in February, dropping its unilateral commitments to pause. CAIS also says the company lobbied to weaken AI safety bills, including California’s SB 1047 and New York’s RAISE Act. It privately backed federal pre-emption of state AI laws, the group adds.
Anthropic had not responded publicly to the statement by Thursday morning. In its amended IPO filing on 10 November, it said it does not expect to safely let Claude fully build its successors. That would need “meaningful human supervision”, it wrote.
The same filing confirmed that Anthropic considers it has met its own threshold for AI-accelerated research . The company had already told investors it could be targeted over safety views.
The statement lands as Anthropic completes the biggest IPO ever. Its shares are due to open on the Nasdaq on Thursday.
Get the most important tech news in your inbox each week.