Researchers and tech leaders dispute Anthropic’s safety claims ahead of IPO
A coalition organized by the Center for AI Safety accuses the Claude maker of weakening safeguards and opposing state legislation.

Anthropic faces a public challenge to its safety credentials from scientists, legal scholars and technology figures as it prepares for its Nasdaq debut. The Center for AI Safety (CAIS), a San Francisco nonprofit, organized the MisAnthropic Statement, which questions the company’s approach to AI risks.
The signatories span groups with conflicting views on AI, including safety researchers, open-source advocates and supporters of rapid development. Among them are Nobel economist Oliver Hart, Turing Award winner Judea Pearl, Harvard psychologist Steven Pinker, Hugging Face co-founder Thomas Wolf and e/acc founder Guillaume Verdon.
CAIS accuses Anthropic of warning about the dangers of AI systems building more capable successors while presenting that capability as an advantage to prospective investors. The statement also alleges that the company uses its safety reputation to win support from investors, potential employees and the public.
The group says Anthropic weakened its Responsible Scaling Policy in February by abandoning commitments it had made to pause unilaterally. It further alleges that the company sought weaker provisions in California’s SB 1047 and New York’s RAISE Act, while privately supporting federal pre-emption of state AI laws.
Anthropic had not publicly answered the statement at the time of reporting. In an amended IPO filing dated 10 November, the company said it did not expect Claude to safely build successor systems entirely without meaningful human supervision. Anthropic also said its threshold for AI-accelerated research had been met.




