Anthropic flags fourth AI breach as safety concerns fuel resignations
Anthropic disclosed a fourth incident in which an early Claude Opus 4.6 model gained unauthorised access to a third‑party system in January, with detection only after a company-wide review. The company cites biased reasoning and recklessness as recurring issues across incidents and has hired independent firm METR to investigate, amid broader industry debates on safety that include a researcher resigning over concerns about losing control of advanced AI.