Anthropic doomer post: '>10%' AI kills all humans

The current debate started with a resignation, not a product launch. Jacob Coxon, a young researcher who has also worked at OpenAI, said he left Anthropic because he believes the leading AI companies are 'gambling with our lives.' Anthropic's alignment lead then amplified that post on X, adding the claim that the chance AI could kill all humans is '>10% within the next decade.' As the Equity podcast panel noted, the combination of a resignation and an internal endorsement made the warning hard to dismiss as routine commentary.

The timing gives the episode its edge. Anthropic is expected to file its S-1 for an IPO within weeks, so the company's own public statements about existential risk sit directly beside the risk factors it must disclose to investors. Sean O'Kane framed the practical question: whether junior lawyers are rewriting that section of the S-1 to say it is officially Anthropic's position that there is a more than 10% chance it could develop something that eradicates humanity and is materially bad for business. Kirsten Korosec raised the opposite possibility, that in the current market such language could function as a valuation flex rather than a warning.

The stakes come down to how the filing handles that tension. If the risk language is already there, the dispute is mostly about emphasis; if it is being added now, the company is deciding in real time how much danger to put in writing. For readers watching the IPO, the test is whether Anthropic's own probability estimate appears in the document and how investors price it.