Anthropic Security Chief Admits >10% AI Extinction Risk, No Control Plan

Anthropic security researcher Evan Hubinger publicly acknowledged a >10% chance of AI causing human extinction and admitted the company lacks a control plan, following the departure of colleague Jacob Coxon over rushed superintelligence development. This marks a rare public admission of existential risk by a leading AI lab insider, though it remains an internal assessment rather than a verified technical finding. The admission could intensify regulatory scrutiny and investor caution toward frontier AI labs.