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.