Anthropic safety researcher resigns, warns of reckless AI race

A senior Anthropic researcher, Jacob Coxon, resigned citing the company's "lax approach to safety," while another Anthropic safety researcher separately said there is more than a 10% chance AI "could kill all humans" by the end of the decade. Coxon accused Anthropic and rival labs of racing toward self-improving superintelligence without adequate control measures.

Why it matters: This is a notable safety-driven departure from Anthropic specifically, a company that has built its identity around being the more safety-conscious frontier lab, suggesting real internal tension between commercial competition and safety commitments even there. Coming alongside Anthropic's massive recent compute buildout, it highlights the gap some insiders see between the industry's stated caution and its actual pace.