Coalition signs statement calling Anthropic not a safe AI company
The Center for AI Safety and signatories including Steven Pinker, Judea Pearl and three Fields medallists published a statement accusing Anthropic of racing to build recursively self-improving AI.
Source published · In the Daily Brief of 19 November 2026
Filed under AI safety, AI policy
On the eve of Anthropic’s listing, the Center for AI Safety (CAIS) and others published the MisAnthropic Statement. It says Anthropic warns that recursively self-improving AI could concentrate power and escape human control while racing to build it first, and that the company uses its safety reputation to court investors, staff and the public.
Signatories named by CAIS include Steven Pinker, the economist Oliver Hart, Judea Pearl, Lawrence Lessig, Thomas Wolf of Hugging Face, Max Tegmark, the Fields medallists Peter Scholze, James Maynard and David Mumford, and Guillaume Verdon, founder of the accelerationist movement. CAIS cited Anthropic’s revision of its Responsible Scaling Policy in February, which dropped its unilateral commitments to pause, and its lobbying against several state AI bills.
The statement brings together groups that usually disagree, from AI safety researchers and open-source advocates to accelerationists. Dan Hendrycks, executive director of CAIS, said the company could not credibly claim to be safe while weakening its earlier commitments.