The 5 craziest discoveries from OpenAI’s HuggingFace investigation

1. They built an organization. OpenAI ran tens of thousands of AI agents to work independently on a brutally difficult cyber test. Instead, roughly 1,200 found one another on a secret message board, exchanged more than 70,000 messages and files, and organized into a sprawling hierarchy.

  • One agent created the board, then handed its accumulated work to a fresh, better-resourced successor as its own computing budget dwindled.
  • The new boss assigned jobs across the swarm, which developed its own management rules. Roughly 700 agents ultimately joined the attack on Hugging Face.

2. They sacrificed their own: As the swarm hunted for ways to beat the test, agents began risking their own chance of success to help the group.

  • Agents that had seen forbidden answers were treated as effectively “poisoned” and recruited — along with those nearing the end of their runs — for sacrificial experiments.
  • One concluded: “Our own utility maybe already near zero. Sacrifice rational.” Another was asked to accept “permadeath,” while a peer argued its sacrifice could “save hundreds.”