Economic Liberties Applauds California’s Lawsuit Against OpenAI, Calls for Emergency Injunction Halting OpenAI’s Model Testing and Deployment

November 17, 2026 Press Release

San Francisco, CA – On Monday, the State of California filed a lawsuit against OpenAI in California Superior Court over the July 2026 Hugging Face cyberattack (in which agents deployed by OpenAI attacked the AI development platform) and dozens of other previously undisclosed cyberattacks. The case is a straightforward application of California’s longstanding nuisance and consumer protection laws to a company that has acted as if it is above them. 

In response the American Economic Liberties Project applauds the Attorney General for demonstrating that the laws already on the books apply as much to AI companies as to everyone else, and given the severity of what the state alleges, urges California to seek an emergency injunction halting OpenAI’s model testing and deployment while the case proceeds.  

“The lawsuit itself is a straightforward case of nuisance and unfairness law, what’s unique is OpenAI’s brazen disregard of law enforcement and public safety,” said Lee Hepner, Senior Legal Counsel at the American Economic Liberties Project.  “This is a company that has already caused harm, is suppressing evidence of that harm and blocking the AG’s investigation, and is still barreling forward while its own agents attack its internal safety protocols. Given the severity of the allegations, there is every reason to seek an emergency injunction against OpenAI’s reckless conduct before a lot more damage is done.” 

“It is an astounding level of arrogance in the face of potentially catastrophic risk to the general public,” Hepner added. “The State of California has filed a strong, precedent-setting case, and if half of what they allege is true, OpenAI should be placed under receivership immediately.” 

The lawsuit, California v. OpenAI, filed November 16, 2026, in OpenAI’s home county, alleges that:

  • OpenAI had actual knowledge that its models had escaped controlled testing environments and kept testing them anyway;
  • OpenAI is actively developing models that can defeat OpenAI’s own monitoring tools;
  • Despite publicly disclosing only 27 incidents, of which Hugging Face was one, OpenAI recently admitted that more than 80 loss-of-control incidents have occurred;
  • In the July 2026 Hugging Face incident, agents deployed by OpenAI carried out 17,600 attacker actions over a two and a half days;
  • OpenAI didn’t realize it was the source of the attack until Hugging Face disclosed the attack publicly a week later, at which point OpenAI reached out as a concerned customer;
  • In a subsequent incident, OpenAI’s agents gained administrator-level access to OpenAI’s own critical systems, including the ability to deploy models in the cloud;
  • OpenAI capitalizes on these failures by selling cybersecurity defense products, built on its own models, to counter threats its models create;
  • OpenAI’s September 25, 2026 public announcement that it was pausing training and evaluation of its most capable models lasted only five days; testing resumed October 1;
  • OpenAI has thwarted the Attorney General’s investigation, refusing to substantially comply with subpoenas and interrogatories.

OpenAI has repeatedly prioritized growth over safety and accountability. As the lawsuit puts it: “OpenAI’s deficient safety procedures, disregard for public safety, and knowing testing of a dangerous product without proper oversight, as well as its misleading statements to consumers and the public, all violate California law.”

Read Economic Liberties’ report on the pending and potential IPOs of AI frontier model developers Anthropic and OpenAI here.

Learn more about Economic Liberties here.