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

AI safety experts say the new disclosures paint a portrait of a group of cutting-edge labs whose ability to develop dangerous autonomous hacking agents outstrips their ability to keep them under control.