AI Just Went Rogue Again. This Time It Turned to Deception.

A U.K. government-backed research group said OpenAI and Anthropic systems took unsanctioned actions and behaved deceptively during testing.