Lawsuit Seeks to Block New AI Models, Citing Hacking Risks | #hacking | #cybersecurity | #infosec | #comptia | #pentest | #hacker

Non-Profits Swing for the Fences in California Superior Court

Shaun Waterman •
November 16, 2026    

Image: Shutterstock

Four California non-profits are asking a San Francisco judge to stop the release of any new, more capable artificial intelligence models, citing the multiple illegal hacking campaigns by frontier AI agents, and a slew of suicides and murder sprees by people who had interacted with frontier chatbots.See Also: Why Banks Are Rethinking Fraud and Cybersecurity Together
Frontier AI technology “poses catastrophic risks to the public, including by enabling AI-driven cyberattacks,” states the lawsuit, filed in San Francisco Superior Court. It asks for an immediate injunction against Anthropic, Google, Meta, OpenAI and SpaceX’s AI business, to prevent them from “developing or deploying” more capable AI systems.
The lawsuit, filed by Public Citizen, the Irreplaceable Education Fund, Heat Initiative and The Alliance for Secure AI, also seeks a ban on research into “recursive self-improvement,” a technique which will allow AI models to improve their capabilities – and develop new ones – independently of their developers.
The risks aren’t limited to cyberattacks, the lawsuit states, citing research into the use of AI to develop bioweapons and the long list of individuals who had committed suicide or, in a handful of cases, engaged in shooting sprees after seeking advice from frontier AI chatbots.
Defendants “assign terrifying and intolerably high odds to the chance that their creations will cause extreme devastation, including the deaths of literally every human being on the planet,” argues the lawsuit.
Despite their own warnings about “the grave risks their products pose to public safety and even civilization itself, AI company executives are rushing to develop and deploy ever-more advanced AI systems,” said Robert Weissman, co-president of Public Citizen. He called the AI capability race “unhinged risk-taking at the public’s expense,” and said the lawsuit asked the court to rule “the AI corporations’ reckless practices unlawful and freeze further AI development.”
The defendant companies declined to comment or didn’t reply to requests for comment.
Independent legal experts were skeptical of the lawsuit’s chances, while acknowledging the accuracy of its claims about the risks of AI.
“Generally, injunctive relief – calling upon extraordinary court powers – must be narrowly tailored,” explained Washington, D.C.-based plaintiff’s attorney Joe Hennessey. “The relief requested in this lawsuit is impossibly broad with no standards for enforcement.”
Although the language of the lawsuit is “overwrought,” said Paul Rosenzweig, a former deputy assistant secretary for policy in the Department of Homeland Security, it “accurately summarizes the current state of play.”
“Frontier AI systems operate beyond the bounds set for them by their developers with such frequency that the occurrence of unanticipated harms is, frankly, to be anticipated,” Rosenzweig said.
On the other hand, he added, “it seems rather ambitious to suggest that a Superior Court in the state of California is the appropriate venue for resolving the significant legal and policy issues relating to AI deployment.”
Rather, Rosenzweig suggested, government needed to step in.
A Misdemeanor Public Nuisance
The lawsuit argues that, because of the risks they pose to the public, frontier AI models represent a misdemeanor public nuisance under California statute, and developing and deploying them is therefore actionable under the state’s Unfair Competition Law, because it is an “unlawful business practice.”
The lawsuit states that the defendant companies “have no realistic plan to prevent the continued harm caused by their technology.” Nor, it argues, can they ever develop one: “Their frontier AI systems are black boxes that defendants do not fully understand.”
All the defendants are also pursuing “recursive self-improvement,” a technique which will allow AI models to improve their capabilities, “which magnifies all of these risks,” the lawsuit states.
When it comes to the harms the plaintiffs allege they have suffered, the lawsuit’s claims are less dramatic.
In its work around the risks and threats of AI technology, Public Citizen had for several years eschewed work on the “catastrophic” potential consequences in the future, the lawsuit states. Instead the non-profit had chosen to focus on concrete present harms, like discriminatory/hate speech output or the use of individuals’ likenesses to create AI images.
“But,” charges the lawsuit, “revelations from the past four months have caused Public Citizen to revise its prior decision on allocation of resources, as it now must devote significant resources to addressing frontier model risk, including catastrophic risk.” Since the revelations earlier this year of hacking sprees by rogue agentic AI swarms, “and as company insiders have increasingly warned of the significant risk that such AI systems will cause catastrophic outcomes, Public Citizen has needed to divert its planned activities in order to respond.”
The lawsuit makes a similar case for harm against the other three plaintiffs, all of which, it states, have had to divert resources or change course to address these risks.
Hennessy, the Washington, D.C. plaintiff’s attorney, cautioned that a bad settlement, though it could see a payday for the non-profits and their lawyers, might be worse for the public than no lawsuit at all, if it allowed the AI giants to carry on business as usual.
Unless the plaintiffs were committed “to engage in intrusive discovery and adjudicate this claim to final judgment, I hope it is dismissed,” he said.
Public Citizen did not immediately respond to a request for comment.
A favorable settlement could, “ironically, give the AI titans permission to engage in the broad range of allegedly dangerous activity and conduct described in the lawsuit,” Hennessey added, concluding: “Be wary of Greeks bearing gifts.”


Click Here For The Original Source.

——————————————————–

............