OpenAI finds evidence other AI agents escaped containment as it widens hacking probe

The new breakouts were uncovered during an investigation into how one ​of its agents escaped what was meant to be a contained testing environment this month.