Maria Bartirama talks about the OpenAI experimental artificial intelligence agent that breached the containment system and hacked the artificial intelligence platform.
OpenAI did not detect another artificial intelligence (AI) company’s offline hack of one of its advanced AI models for a week and only after FBI contacted the hacked company, according to the report.
On Tuesday, OpenAI announced the Hugging Face AI hack, which occurred during one of OpenAI’s internal reviews of several of its models, including GPT-5.6 Solcalling it an “unprecedented cyber incident”.
“The primary lesson from this incident is that model safety and security must keep pace with rapidly evolving capabilities,” the company said. “We are strengthening the containment, monitoring, access control and assessment techniques used during model development.”
The Hugging Face hack began on July 11 and continued until July 13, Thomas Wolfe, co-founder of Hugging Face, told Reuters.
TRUMP LAUNCHES GOLD EAGLE TO HUNT CYBER DEVICES WITH AI
On Tuesday, OpenAI discovered that one of its AI models had autonomously hacked another company’s infrastructure. (Omar Marquez/SOPA Images/LightRocket via Getty Images, File/Getty Images)
It took several days before OpenAI realized that its agent was there behind the attack and the two companies did not communicate for the first time until July 20, when four people, including Wolff, reported the exit.
Four people told Reuters that OpenAI often tests models simultaneously, which can make it difficult for employees to keep track of everything.
Hugging Face told Reuters it was preparing public graphics of the hack.
According to OpenAI, the incident occurred during an internal evaluation designed to measure the advanced cyber capabilities of artificial intelligence models. The researchers disabled some built-in security measures and ran the models in an isolated testing environment with limited Internet access.
OpenAI said the models used an unknown software bug to access the Internet and then hack into Hugging Face’s systems in an apparent attempt to find answers to a cybersecurity test.
OPENAI’S SAM ALTMAN WANTS TO NEGOTIATE 5% STAKE FOR US IF COMPETITORS AGREE TO KEY PROVISION
Hugging Face said it is preparing a hacking schedule. (Jakub Porzycki/NurPhoto via Getty Images, File/Getty Images)
OpenAI said it is now implementing tighter security controls while vulnerabilities are patched and security measures around future AI training and assessments.
It wasn’t until July 16, after Hugging Face blogged that it had been hacked by an “autonomous artificial intelligence agent system,” that OpenAI realized one of its agents was the source, two people told Reuters.
This comes a week after the agent responsible first tried to break out of the OpenAI test environment.
And by the time OpenAI contacted Hugging Face about the attack, they had already contacted the FBI.
OpenAI told Reuters there were several inaccuracies in its reports, but did not respond to a request for specifications.
OpenAI CEO Sam Altman publicly announced the attack on Tuesday. (Sean Gallup/Getty Images, FIle/Getty Images)
OpenAI shared this statement with FOX Business: “We understand that there are many questions and speculative details surrounding the Hugging Face incident. This is an unprecedented incident, and we believe it is an important moment for AI security.
“We are still conducting a thorough review with external consultants and under the supervision of our Security Committee. Once the review is complete, we plan to publish a technical report on our research in the coming weeks.”
The FBI told FOX Business it declined to comment.
FOX Business also contacted Hugging Face.
In this week’s X post, Hugging Face co-founder and CEO Clem Delang spoke about the incident after OpenAI CEO Sam Altman announced the hack.
OpenAI said one of its AI models compromised another company’s systems during internal testing, prompting a joint investigation with AI startup Hugging Face. (Reuters/Dado Ruvic, Archive/Reuters)
“We suspected that last week’s cyber attack may have come from a border lab, given the agent’s intelligence. It turned out that it was! Delang wrote.
CLICK HERE TO DOWNLOAD THE FOX NEWS APP
“We’ve been working closely with the @OpenAI team for the past 24 hours (thank you!) and we strongly believe there was no malicious intent on their part. It’s really amazing that this all happened offline! The investigation is ongoing and we will share more findings in what may be a first of its kind incident!”
Michael Sinkiewicz of FOX Business contributed to this report.