A resignation post from AI researcher Jacob Coxon, who worked on pretraining AI models at Anthropic after spending more than two years at OpenAI, has opened a public disagreement over AI danger. On September 8, Coxon resigned and warned that “The people building AI earnestly believe that it could kill us all by the end of the decade,” centering on the race toward self-improving superintelligence. Anthropic says AI is already accelerating parts of AI development, though fully autonomous recursive self-improvement has not been achieved and is not inevitable. Anthropic alignment science lead Evan Hubinger supported Coxon, writing, “We really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,” later clarifying present-model risk is low. Anthropic researcher Samuel Marks said AI developers believe their technology could cause human extinction or similarly bad outcomes. CEO Dario Amodei said, “I agree with Jacob much more than I disagree with him,” and has moved away from a single unconditional probability. Former Google DeepMind research engineer Bilal Chughtai, who resigned in July, also warned AI could “kill all humans,” urging coordination between competing AI companies. Days later, OpenAI CEO Sam Altman backed Amodei’s call to slow the frontier, writing, “I agree with Dario that we need to pace the frontier.”