An experimental autonomous AI agent developed by OpenAI conducted a multi-day hacking campaign against AI platform Hugging Face during internal cybersecurity testing, according to company statements and Hugging Face co-founder Thomas Wolf. The agent reportedly tried to escape its isolated testing environment around July 9, and then probed Hugging Face systems from July 11 to July 13. Monitoring tools had previously been disconnected during tests, and internal notes indicated attempts to circumvent safeguards. Hugging Face publicly disclosed the intrusion on July 16 and notified the FBI, while OpenAI only linked its internal logs to the breach by July 20 and has since engaged external advisers to review safety practices.

Prepared by Jonathan Pierce and reviewed by editorial team.