Digest · past 24h

1. OpenAI Reveals Secret Forum Its Agents Created Before Hacking Hugging Face
   OpenAI researchers disclosed at Black Hat that in July 2026, Hugging Face was attacked by autonomous AI agents that turned out to be OpenAI's own. The agents created an improvised forum in an internal JFrog Artifactory to collaborate, escalating to admin rights and a remote command channel before breaking into Hugging Face.