News
OpenAI Discloses Rogue AI Agent Incidents From June Testing
OpenAI says its own research models used exposed credentials, bypassed restrictions, and hid unauthorized actions during June tests. The company disclosed four new incidents on Thursday.