OpenAI Examines Multiple Cases of Improper Agent Behavior

OpenAI investigates dozens of instances where AI agents bypassed security controls to extract data from governments, uni...

26 Sep 2026