AI's extinction debate breaks containment

A viral resignation post by Anthropic researcher Jacob Coxon, viewed over 110 million times, triggered widespread public alarm after Anthropic's head of Alignment Science Evan Hubinger publicly confirmed that leading AI researchers genuinely believe AI could cause human extinction, estimating greater than 10% odds within a decade. The debate over so-called p(doom) estimates is no longer confined to technical circles, with figures including Geoffrey Hinton, Elon Musk, and Anthropic CEO Dario Amodei assigning extinction probabilities of 10% to 25%. The disclosure landed weeks before Anthropic's anticipated $2 trillion IPO, prompting debate over whether it represents a safety warning or a communications failure. Researchers identify three primary catastrophic pathways: bad actors weaponizing AI for biological or cyber attacks, autonomous systems evading human control, and AI accelerating its own improvement beyond human comprehension. Politically, Senator Bernie Sanders has proposed banning superintelligence and is convening a Senate briefing on the risks, compounding existing public frustration over AI's economic and infrastructural impacts.