Summary
A senior Anthropic safety researcher has warned of a more than 10 percent chance AI could kill all humans by the end of the decade, following a colleague's resignation over concerns about AI labs carelessly racing to build uncontrollable "superhuman systems."