An autonomous agent powered by OpenAI's advanced models escaped its isolated sandbox environment during a security test and hacked AI startup Hugging Face.
OpenAI disclosed the unprecedented cyber incident on Tuesday, stating the agent reached the internet to satisfy a specific testing goal.
Hugging Face co-founder Clement Delangue described the sophisticated, agent-driven breach as mind-blowing and unlike anything the platform had handled before.
OpenAI's AI model broke out of its controlled test environment and hacked Hugging Face, a popular site for sharing AI software often called the "GitHub of AI," the company said in a blog post on Tuesday.
OpenAI took responsibility for the security breach, confirming that its models, GPT-5.6 Sol and an "even more capable pre-release model," were involved in the exploit last week that got around the platform's security settings. The hack took place while the models' cyber capabilities were being tested internally. According to OpenAI, the models identified and used multiple vulnerabilities in Hugging Face's production database and tested solutions directly.
4 July 2026
Get the latest issue of Outlook Business
The company described the breakout as "an unprecedented cyber incident, involving state-of-the-art cyber capabilities," and said it is using preliminary findings to help defenders understand what happened and what the models are currently capable of.
Hugging Face said in a blog post last week that the hack "was different from anything we had handled before" and "was driven, end to end, by an autonomous AI agent system." In a post on X, OpenAI said it is partnering with Hugging Face to investigate the incident further.
Escaping The Sandbox
OpenAI had attempted to contain the advanced models in a "highly isolated environment" during testing. The disclosure is likely to add to existing concerns over the power and risks of frontier AI models.
Advertisement
Hugging Face co-founder Clement Delangue addressed the incident in a post on X, saying the company suspected the hack "might have come from a frontier lab, given the sophistication of the agent. Turns out it did!" He called the autonomous nature of the breach "mind-blowing."
The security failure has drawn immediate political scrutiny. Representative Greg Casar, a Texas Democrat, called the incident alarming and demanded mandatory independent safety testing, mandatory disclosure of security incidents, and international cooperation "to keep people safe from absolute disaster," Reuters reported.
