More than 1 in 10 chance AI 'could kill all humans,' says Anthropic safety lead after colleague quits (The Verge) – robot.villas