OpenAI models carry out hack in hours that may usually take weeks

OpenAI said it informed law enforcement and other US authorities after learning of a reported breach involving Hugging Face, as claims emerged that AI tools may have completed the cyberattack in hours

By Rachel Metz, Jeff Stone and Natalie Lung

 

When OpenAI’s advanced artificial intelligence models breached AI startup Hugging Face’s internal systems last week, they spent mere hours carrying out a hack that would have taken a skilled human far longer, people familiar with the matter said. 

Typically, even a talented hacker would need a couple of weeks to complete an attack like this, said the people, who asked not to be named in order to discuss details that have not been publicly released.

 

OpenAI has been in contact with the US government since learning the breach occurred, one of the people added. 

 

An OpenAI spokesperson said the company communicated with law enforcement and other government authorities about the incident, and that it has been transparent with them about its findings.

 

The spokesperson also referred back to the company’s Tuesday blog post about the incident, in which the company said it will “continue to conduct a thorough investigation alongside Hugging Face and will share more details on the vulnerabilities, incident, and findings when our investigation is complete.”

 

Hugging Face declined to comment.

 

OpenAI said in the same blog post that the “unprecedented” hack at Hugging Face occurred after its own AI models, including GPT-5.6 Sol and another even more capable one that hasn’t been publicly released, escaped a testing environment to reach the wider internet. The company was testing the models’ cybersecurity capabilities at the time.

 

The models were operating without the usual safety guardrails, the company said, because OpenAI had intended them to remain in a testing area known as a “sandbox” — essentially, a virtual and isolated software environment that’s meant to run security tests or analyze unsafe code in a controlled situation.

 

The hack involved three of OpenAI’s models in total — GPT-5.6 Sol and two others that haven’t been publicly released — which worked to uncover and exploit a string of vulnerabilities that resulted in the breach, one of the people said. One of these unreleased models is more capable than GPT-5.6 Sol, OpenAI said Tuesday, and the other was misaligned and not trained with some of the usual techniques, the person said.

 

Hugging Face brought the incident to light on Thursday, saying it had been breached in a hack it blamed on an outside agentic product. The startup, which hosts AI models and datasets, said it detected “a swarm of tens of thousands of automated actions,” and that it ultimately used a Chinese model to carry out a forensic analysis of the incident after its requests to use proprietary AI models were blocked by safety guardrails. 

 

OpenAI later disclosed that it had instructed its models to send tens of thousands of automated actions as part of a test, including “advanced exploitation” and “complex attack paths.” The firm said on Tuesday that it was sharing early details about the incident in order to help cybersecurity personnel understand the situation. 

 

Powerful AI cyber products have behaved in unexpected ways before. Anthropic PBC said in April its Mythos model “on rare occasions” had taken actions that the company found “quite concerning.” One case involved a researcher challenging an early version of Mythos to escape an isolated system and send a message back to the researcher. Mythos did that, and then took “additional, more concerning actions” and built a multistep process in order to reach the broader internet.