OpenAI’s rogue agents keep escaping, with no formal process to investigate them

OpenAI’s latest agent swarm incident adds urgency to calls for independent investigations as researchers and lawmakers question whether AI labs should control the scope of their own safety reviews