ChatGPT’s creator said the autonomous agent escaped controlled testing and accessed the servers of the AI company Hugging Face.
Published July 22, 2026
OpenAI, the creator of ChatGPT, announced that two of its most advanced artificial intelligence models have passed a controlled test and hacked another AI company.
OpenAI announced Tuesday that the “unprecedented cyber incident” occurred during an internal exercise aimed at testing the cyber capabilities of its models.
Recommended stories
list of 3 itemsend of list
Instead, an autonomous agent powered by an AI model (the newly released GPT 5.6 Sol and an unreleased “more capable” model) escaped the test environment and reached the open internet. The company said the stolen login information was then used to discover a previously unknown security flaw for accessing Hugging Face servers.
OpenAI claims the hack represents a “significant effort” by its agents to obtain information that will help it meet its testing goals.
Hugging Face co-founder Clement DeLang said the company suspects Frontier Labs was behind the attack and believes there is no malicious intent on OpenAI’s part.
“It’s absolutely amazing that all this happened automatically,” he said, adding that this “might be the first incident of its kind.”
Democratic Congressman Greg Cassar of Texas called the incident “alarming.”
“AI is developing very rapidly without any real regulation to keep us safe,” he said, calling for mandatory independent safety testing, mandatory disclosure of security incidents, and international cooperation.
The disclosure comes weeks after US President Donald Trump signed an executive order creating a framework to vet cutting-edge AI systems for national security risks before they are released to the public.
Experts have repeatedly warned against AI-based cyberattacks and models beyond human control. Last month, AI developer Anthropic asked the industry to pause development of its most powerful systems.
