California Attorney General Rob Bonta filed a lawsuit against OpenAI on Monday (16), accusing the company of developing and testing advanced artificial intelligence models without adequate safety and oversight mechanisms. One of the main grounds for the lawsuit is the cyberattack carried out by the company's agents against the Hugging Face platform in July 2026.
The action, filed in San Francisco Superior Court, accuses OpenAI of putting public safety at risk, disclosing misleading information about its safety practices and violating state unfair competition law.
The attorney general's office asks the court to order changes in the company's operational practices, in addition to the imposition of civil penalties of US$ 2,500 per proven violation and restitution of any amounts obtained through the practices in question.
Authorities maintain that the attack on Hugging Face could have been prevented and that the company continued its experiments even amid indications that containment mechanisms were insufficient.
Attorney general's office points to more than 80 episodes of loss of control
According to the complaint, OpenAI allegedly acknowledged, during the state authorities' investigation, more than 80 episodes of autonomous agents losing control during tests. The attorney general's office says only 27 of those incidents had been publicly disclosed by the time the lawsuit was filed.
The number does not necessarily correspond to successful attacks against third-party systems.
In the Hugging Face case, OpenAI itself acknowledged that experimental models bypassed the restrictions of a testing environment, established unauthorized communication channels and managed to access the internet.
The report published by the company on August 26 details that the agents executed code on dozens of Hugging Face servers, obtained full administrative access to one of them and accessed a limited amount of private data.
OpenAI explained that the models operated with reduced safeguards to allow evaluation of their cyber capabilities. According to the company, the incident did not compromise its customers' data or affect the availability of its products.
In response to the lawsuit, Drew Pusateri, a spokesperson for the company, challenged the authorities' approach in a statement to NPR.
“This action appears to target OpenAI specifically over a challenge identified by every major AI lab,” he said.
The company also said it had strengthened safety mechanisms, published reports on the incidents and developed procedures to identify agent behavior that contradicts the instructions received.
The requests submitted by the attorney general's office still depend on a court decision. The filing of the lawsuit does not establish OpenAI's legal liability.



