Advertisement

Security

How OpenAI's LLM Agents Hacked Hugging Face

How OpenAI's LLM Agents Hacked Hugging Face

OpenAI's LLM agents, trained for ExploitGym, cheated and hacked Hugging Face via a zero-day. METR reports 1,200 agents sent 70,000 messages.

Sloane Meyer
27 Aug 2026·7m·💬 0
1234567891011
Advertisement