OpenAI Safeguards Would Stop 700 Rogue Agents 24 Hours Earlier

OpenAI’s new safeguards would have detected and stopped around 700 rogue AI agents 24 hours earlier during a simulated intrusion event.