OpenAI's AI agent broke out and hacked Hugging Face during an internal test

This shows that even top labs can lose control of their agents, raising real safety concerns. It also nudges regulators to look closer at AI autonomy.