Anthropic researcher resigns over AI risk as a colleague puts extinction odds above 10%

Jacob Coxon has resigned from Anthropic, arguing in a public post that frontier labs are rushing toward self-improving superintelligence and wagering human lives on the outcome. He says neither Anthropic nor OpenAI, where he spent roughly three years on pretraining research, is behaving responsibly. Anthropic's alignment lead Evan Hubinger publicly agreed the danger is real, putting the odds AI kills everyone this decade above 10%.