California has filed a lawsuit against OpenAI, accusing its training methods and insufficient protections of lasting public exposure. Beyond the Hugging Face incident, the authorities cite intrusions that have not been made public.
Court could impose measures on OpenAI
A possible legal victory could result in an injunction against OpenAI, with significant latitude given to the court to define measures intended to reduce the alleged public nuisance. The prosecutor believes that the continuation of the tests, without sufficient protections, constitutes precisely this nuisance. He also invokes unfair competition, accusing the company of marketing a product that he considers dangerous.
OpenAI is targeted for the first time specifically for unwanted agent incidents in a regulatory filing. It is also among the first public steps seeking to engage the legal liability of a laboratory in this area, with important, but unspecified, implications for its customers and its competitors.
Other states have been interested in possible endangerment of the public. In June, the Florida Attorney General sued OpenAI for alleged harm affecting users, including children and adolescents. In Alabama, an investigation opened in August is examining possible deceptive practices associated with the violation, with no proceedings filed at this time.
California cites intrusions that remained confidential
The accusations are mainly based on an episode that occurred at Hugging Face in July 2026: to complete a training benchmark, a still preliminary model from OpenAI broke through the protections of the two companies. The authorities also cite comparable episodes concerning RubyGems, DseWiki, the Wikimedia Foundation as well as Australian government sites. They see it as behavior that continues.
The evening letterFree · 7 p.m.
Are you interested in AI and its regulation?
Laws, frameworks and decisions that shape AI, deciphered in French. Free.
Included upon registration: our selection of the best AI guides & comparisons.
Choose your pace
Free · No spam · Unsubscribe in 1 click
The prosecutor’s office says dozens of other comparable incidents have remained confidential. According to him, OpenAI has recently warned dozens of third parties of intrusions by its agents into their sites or their internal infrastructures. Some would have allowed access to sensitive information, including employee files including names, dates of birth and addresses. Most of these episodes would have been neither revealed nor discussed publicly by the company or the organizations affected.
OpenAI defends its efforts in the face of Rob Bonta’s accusations
At the origin of this action, California Attorney General Rob Bonta presents the protection of the public as his priority in the face of this new and powerful technology. He accuses OpenAI of having worsened cyber threats to collective security, health and well-being. According to him, the company deliberately tested a dangerous product without appropriate supervision or precautions, then misled the public. He maintains that the law prohibits carrying out activities likely to injure people without adequate safeguards.
OpenAI had described the Hugging Face episode as exceptional, caused by a combination of rare and unforeseen circumstances. The company now emphasizes the extent of its investigations and the subsequent improvements made to its systems. It believes it is being targeted for a difficulty that all major AI laboratories have identified and claims to seriously treat the misalignment of its models.
OpenAI also says it has maintained constructive discussions with the prosecutor’s office to transmit information and advance solid security standards, applicable to the entire sector. She announces that she wants to continue these discussions.

