OpenAI Launches Months-Long Audit of Rogue Agent Behavior

OpenAI says an 'extensive' review of model agent activity will take months, after disclosing breaches at Hugging Face, Australia's Medicare portal, and several U.S. federal sites.