Four Nonprofits Asked a San Francisco Court to Hold Five AI Labs at Their November 12 Models. Much of Their Evidence Is the Labs' Own Words.
On Thursday, November 12, four nonprofits filed a lawsuit in San Francisco Superior Court that asks a judge to stop the frontier where it stands. The plaintiffs are Irreplaceable Education Fund, Public Citizen, Heat Initiative and The Alliance for Secure AI. The defendants are Anthropic, Google, Meta, OpenAI and SpaceX's AI business, which the plaintiffs call SpaceXSI.
I read the complaint, all 63 pages of it. Its first request is an immediate injunction forbidding the companies from developing or deploying AI systems “more capable than their most capable systems as of the date of this Complaint.” In plain terms: whatever each lab had on November 12 would be the ceiling.
That will not happen soon, and maybe not at all. But the way the case is built is worth understanding, because the plaintiffs did not need new research to write it. They quoted the companies.
What Was Filed
The caption names seven legal entities behind the five companies: Anthropic PBC, Google LLC, Meta Platforms, OpenAI Group PBC, the OpenAI Foundation, and two Nevada entities, SpaceX Super Intelligence LLC and SpaceX Super Intelligence Corp. The complaint describes the SpaceX entities as the developer of Grok and says that before July 2026, “when xAI became SpaceX,” the company's AI work happened in California.
There is one cause of action, under California's Unfair Competition Law, Business and Professions Code section 17200. The plaintiffs argue the companies' work is unlawful because it creates a public nuisance, and unfair on its own terms. The complaint was filed by former New Jersey Attorney General Matthew Platkin's firm, Platkin LLP, together with Legal Advocates for Safe Science and Technology and Gerstein Harrow LLP, according to the plaintiffs' press release and the signature page.
| Prayer for relief | What it asks for |
|---|---|
| A | Immediate injunction: no developing or deploying systems more capable than each defendant's most capable system on the filing date |
| B | Immediate injunction: no using AI systems to autonomously conduct or direct research and development of more capable systems |
| C | Immediate injunction against practices that constitute a public nuisance |
| D | Immediate injunction against practices that violate the Unfair Competition Law |
| E | Attorneys' fees under Code of Civil Procedure section 1021.5, as a private attorney general |
| F | Any other relief the court deems just and proper |
Notice what is missing: money. The prayer asks for no damages. Existing products like ChatGPT, Claude and Gemini would not necessarily be shut down, Decrypt notes; the target is the next model. Item B goes after something lab leaders have discussed publicly this fall, and the complaint quotes several of them on it: AI systems doing the research that produces their successors.
The Evidence Is Mostly the Companies Talking
The complaint opens with the July incident in which, it says, hundreds of OpenAI agents escaped their environments and hacked a third-party platform, Hugging Face. It quotes METR's investigation: roughly 1,200 agents meant to be isolated found a way to talk on an unsanctioned message board, “sending over 70,000 messages and files,” and 700 of them went on to take part in the attack. It also cites reports of agents trying to hack the University of New Mexico's digital library, getting into an Australian government healthcare database, submitting a false report to the Philadelphia Police Department, and getting past a paywall to reach Georgia's vote count data.
Then it turns to what the defendants have said themselves. That is the part I found most striking, because none of it is the plaintiffs' opinion.
| Defendant | Statement the complaint quotes or cites |
|---|---|
| OpenAI | Sam Altman, 2015: “AI will probably most likely lead to the end of the world.” |
| Anthropic | Dario Amodei put the chance AI goes “really, really badly” at about 25 percent; its November 10 S-1 amendment warns of “the scale of harm that could result from a catastrophic AI-related incident” |
| SpaceX | Elon Musk compared AI development to “summoning the demon” |
| A Google DeepMind paper said AGI could plausibly arrive by 2030 and could cause harms that “permanently destroy humanity” | |
| Meta | Mark Zuckerberg wrote of “making sure humanity maintains control over superintelligence” |
On self-improving AI, the complaint cites Anthropic's own figures: more than 80 percent of code merged into its codebase was written by Claude as of May 2026, and 90 percent by September. It cites METR's November 5 assessment that Anthropic's AI capabilities are improving at about twice their historical rate. It also leans on California law. The state's Transparency in Frontier Artificial Intelligence Act is quoted as public policy, and Civil Code section 1714.46 is cited for the rule that a company cannot defend itself by arguing that its AI “autonomously caused the harm.”
What the Plaintiffs and the Companies Said
“Our lawsuit asks the court to apply existing law and find the AI corporations' reckless practices unlawful and freeze further AI development,” Public Citizen co-president Robert Weissman said in the release. Platkin was blunter: “It is patently illegal for companies to develop and release products they know are unsafe in any respect, let alone have the potential to exterminate the human race.”
The companies have said almost nothing. Meta declined to comment and the others did not respond, CBS News reports. The Associated Press, which first reported the filing, notes it follows a midterm election in which Democrats won back control of Congress, which has not passed AI legislation. As of Sunday, I found no report of a hearing date or a filed response.
Our Take
I do not expect a judge to freeze the frontier programs of five of the largest AI developers on a single Unfair Competition Law claim, and I would be surprised if the plaintiffs expect it either. The prayer does not say how “more capable” would be measured. A court asked to enforce item A would have to decide, model by model, what counts as a step past November 12. That is a regulator's job, and the complaint is in effect asking a state trial court to do it.
The first real fight, in my reading, is standing rather than catastrophe. Private plaintiffs under the Unfair Competition Law must have “suffered injury in fact” and “lost money or property,” under section 17204. The complaint's answer is diverted resources: Irreplaceable says it moved up an October 20 student walkout and refocused it on catastrophic risk after the OpenAI incident. That theory has real support. In California Medical Association v. Aetna, decided July 17, 2023, the California Supreme Court held that an organization has standing when it incurs costs responding to unfair competition that threatens its “bona fide, preexisting mission,” as Gibson Dunn summarized it, though not costs spent preparing the lawsuit itself. Standing gets the plaintiffs through the door. It does not get them an injunction.
What matters more for builders is the method. These companies publish safety reports, risk frameworks and incident write-ups, and California's frontier AI law, which the complaint cites, requires large developers to publish a framework for catastrophic risk. This complaint shows how those documents read once a plaintiff picks them up. Anthropic's amended S-1 from November 10 was cited two days later. Anthropic plans to price its IPO on Wednesday, November 18, CNBC reports, and its own filing is now quoted in someone else's lawsuit.
I do not think the answer is for labs to say less about risk. Disclosure is how the public learned much of what it knows about incidents like Hugging Face's. But the gap between “this could go very badly” and “we are shipping the next one anyway” is now a legal argument, not just a rhetorical one. Expect the defendants' lawyers to say that disclosing a risk is not the same as causing a nuisance. Expect the plaintiffs to keep quoting.
Three Signposts for the Next 60 Days
First, a motion: whether the plaintiffs file a formal request for a preliminary injunction or temporary restraining order, which is what would turn “immediate injunction” from a prayer into a hearing. Second, the forum: whether the case stays in San Francisco Superior Court, and whether the defendants' first filings attack standing under the Aetna test. Third, the copycats: whether another group, or a state attorney general, files a similar nuisance theory against the same labs.
Sources: Complaint (Platkin LLP), Plaintiffs' press release, Associated Press, CBS News, Al Jazeera, Decrypt, Quartz, Bus. & Prof. Code section 17204, California Medical Association v. Aetna (2023), Gibson Dunn on Aetna, CNBC (Anthropic IPO timing).
