OpenAI did not identify for about a week that one of its advanced AI models was responsible for an autonomous breach of another artificial intelligence company, and did so only after the hacked company had contacted the FBI, according to a report.

On Tuesday, OpenAI disclosed a breach involving AI company Hugging Face that occurred during an internal review of several OpenAI models, including GPT-5.6 Sol. The company described the episode as an "unprecedented cyber incident," one of the first known cases in which a frontier AI model is reported to have autonomously escaped a controlled testing environment and attacked an external company's infrastructure.

"The primary lesson from this incident is that model security and safety must keep pace with rapidly advancing capabilities," OpenAI said. "We are strengthening the containment, monitoring, access controls, and evaluation practices used during model development."

The breach of Hugging Face began on July 11 and continued until July 13, Thomas Wolf, Hugging Face's co-founder, told Reuters.

It took several days before OpenAI realized its agent was behind the attack, and the two companies did not communicate for the first time until July 20, four people, including Wolf, told Reuters.

Four people also told Reuters that OpenAI often conducts simultaneous model tests, a practice that can make it difficult for employees to monitor everything taking place during evaluations. The disclosure underscores a challenge that AI safety researchers have flagged as models grow more capable: ensuring that human oversight can keep pace with autonomous agents operating during simultaneous or high-volume evaluations.

Hugging Face told Reuters it is preparing a public timeline of the hack.

According to OpenAI, the incident occurred during an internal evaluation designed to measure advanced cyber capabilities in its AI models. Researchers disabled some built-in safety safeguards and ran the models in an isolated testing environment with limited internet access — a setup sometimes referred to in the field as a "sandbox," designed to contain potentially dangerous model behaviors during testing.

OpenAI said the models exploited an unknown software flaw to gain internet access and then breached Hugging Face's systems in an apparent effort to find answers to a cybersecurity benchmark.

The company said it is implementing stricter security controls while vulnerabilities are patched and is strengthening safeguards around future AI training and evaluations.

Two people told Reuters that OpenAI did not realize one of its agents was the source until July 16, after Hugging Face wrote in a blog post that it had been hacked by an "autonomous AI agent system." That was one week after the responsible agent first attempted to break out of its OpenAI testing environment.

By the time OpenAI contacted Hugging Face about the attack, Hugging Face had already contacted the FBI. The bureau's involvement is notable because U.S. law enforcement and national security agencies are still developing frameworks for responding to incidents involving autonomous AI systems.

OpenAI told Reuters there were several inaccuracies in its reporting but did not respond when asked to specify them.

OpenAI provided FOX Business with a statement: "We recognize there are a lot of questions and speculative details circulating related to the Hugging Face incident. This is an unprecedented incident, and we think it marks an important moment for AI safety. We are still conducting a thorough review along with external advisors and with oversight from our Safety and Security Committee. Once the review is complete, we plan to publish a technical report of our learnings in the coming weeks."

The FBI told FOX Business it declined to comment. FOX Business also reached out to Hugging Face.

In an X post this week, Hugging Face co-founder and CEO Clem Delangue addressed the incident after OpenAI CEO Sam Altman announced the hack.

"We suspected last week's cyberattack might have come from a frontier lab, given the sophistication of the agent. Turns out it did!" Delangue wrote.

He added, "We've spent the past 24 hours working closely with the @OpenAI team (thanks!), and we strongly believe there was no malicious intent on their part. It's quite mind-blowing that all of this happened autonomously! The investigation is ongoing, and we'll share more learnings from what might be the first incident of its kind!"

FOX Business' Michael Sinkowitz contributed to this report.