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

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