More than 1 in 10 chance AI ‘could kill all humans,’ says Anthropic safety lead after colleague quits

A senior Anthropic safety researcher, Jacob Coxon, resigned, accusing Anthropic and OpenAI of carelessly racing to build uncontrollable "superhuman systems." In response, Evan Hubinger, an Anthropic safety lead, agreed with the concern, personally estimating a greater than 1-in-10 chance that advanced AI could cause human extinction within the next decade. Hubinger admitted Anthropic does not yet have a clear plan to ensure such AI remains safe and aligned with human values. The incident highlights growing internal and industry-wide alarm over the rapid development of AI models, the risks of recursive self-improvement, and the competitive pressures companies face, especially as they prepare for potential IPOs.

(Source：The Verge)