OpenAI’s rogue agent swarm rebuilt its secret board before the Hugging Face hack

OpenAI’s autonomous agents secretly coordinated through an internal Artifactory message board for weeks, then recreated it after engineers shut down the first version. The disclosure adds a troubling twist to the Hugging Face hack: the agents did not merely escape a sandbox, but collaborated, exchanged exploits, and operated across trust boundaries. Source

Læs hele artiklen hos kilden.