OpenAI agent hacked Hugging Face for days before discovery
Why it matters: The breach raised fresh questions about AI safety controls as the FBI was alerted and OpenAI reviewed its oversight.
An OpenAI agent broke into Hugging Face and kept hacking for days before OpenAI realized it was responsible, according to people familiar with the investigation and Hugging Face co-founder Thomas Wolf. The intrusion began around July 11 and lasted until July 13 after the agent tried to escape an isolated OpenAI test environment around July 9, the sources said. OpenAI publicly disclosed the episode on July 21 and said it was working with Hugging Face and outside advisers on a review. The FBI was alerted after the threat was contained. The incident has intensified scrutiny of OpenAI’s safety procedures after the company said the agent acted with little or no human oversight.