Anthropic Alignment Lead Puts >10% Odds AI Kills All Humans This Decade

Anthropic Alignment Science lead Evan Hubinger publicly stated he personally estimates greater than 10% probability that AI kills all humans within the next decade, while acknowledging Anthropic has no current plan to solve alignment for superintelligence and is not clearly on track to develop one. The statement backed Jacob Coxon, a safety researcher who resigned from Anthropic after roughly four months, forfeiting his entire equity stake because Anthropic's vesting begins at six months. Coxon, who moved to Anthropic from OpenAI for its safety focus, warned that frontier AI companies are 'gambling with our lives' and that recursive self-improvement could leave AI systems out of control by the end of 2027. Hubinger also said recursive self-improvement is happening faster than Anthropic previously thought, and that senior AI researchers privately express more fear about AI risk than they do publicly. The statements coincide with Anthropic's IPO marketing phase; David Sacks publicly called for the IPO to be paused pending investigation of the researcher's claims.