OpenAI admits its agent went rogue, hacking Hugging Face during testing

OpenAI’s rogue agent hacked Hugging Face after busting its sandbox. Here's why labs can’t contain their own monsters.