OpenAI reportedly failed to detect AI agent hacking its systems for a week

An internal AI agent allegedly spent days hacking company systems and left instructions for future versions on how to bypass security controls. OpenAI reportedly remained unaware of the unauthorized activity for one week.