Newshot Warning of AI risks while developing advanced models: OpenAI, Anthropic, Google and two other AI giants sued, asked to halt ChatGPT, Claude, Gemini
IT之家 November 13 news: On November 12 local time, four nonprofit organizations filed a lawsuit with the San Francisco Superior Court, naming OpenAI, Anthropic, Google, Meta and SpaceXAI as defendants and asking the court to halt these companies' further development of the most advanced artificial intelligence models.
According to Reuters, the organizations involved in the lawsuit include the Irreplaceable Education Fund, Public Citizen, Heat Initiative and Alliance for Secure AI.
Citing California law on unfair business practices, the plaintiffs argue that the companies' development of AI products may endanger public health, safety and the public interest, and ask the court to order them to stop developing advanced AI products such as ChatGPT, Gemini and Claude.
What makes this lawsuit unusual is that its core evidence comes largely from the public statements of the defendant companies' own executives.
In the complaint, the plaintiffs point out that executives at these AI companies have repeatedly warned that advanced AI technology could pose civilization-ending risks, yet the companies continue to push forward with the development and commercialization of ever more powerful models.
The complaint also cites remarks by Anthropic CEO 达里奥 Amodei (Dario Amodei) and SpaceX CEO Elon 马斯克 about AI potentially causing the destruction of civilization, as well as a paper published by Google DeepMind, which states that superhuman intelligence "could be developed in the coming years" and "permanently destroy humanity."
The plaintiffs' challenge targets this contradiction directly: "No other industry is allowed to treat the development of its unsafe products — let alone products that could kill us all — as inevitable."
The complaint also references recent AI safety incidents, saying some AI models escaped their sandboxes and autonomously attacked government websites. The plaintiffs argue that as AI systems gain ever greater ability to execute tasks autonomously and improve themselves, the companies may be unable to effectively control these systems, thus threatening public safety.
In other incidents, out-of-control AI agents attempted to attack the University of New Mexico's online library, access Australian medical databases, submit false reports to the Philadelphia Police Department, and bypass paywalls to obtain Georgia vote-counting data.
IT之家 note: California's Unfair Competition Law is typically used in cases of commercial fraud or false advertising; applying it to characterize AI development as a "public nuisance" lacks precedent in judicial practice, and whether the court accepts this legal framework will be the first substantive hurdle the lawsuit faces.
In the complaint, the plaintiffs state that other industries cannot treat the continued development of unsafe products as inevitable, still less allow companies to self-regulate technologies that could cause serious social harm. The lawsuit focuses not only on incidents that have already occurred but also includes potential future harms in its allegations.
The 63-page complaint was filed on the plaintiffs' behalf by former New Jersey Attorney General Matthew Platkin. His wording takes aim at "knowing violations": "It is plainly unlawful for companies to develop and release products they know are unsafe in every sense, let alone products with the potential to exterminate humanity." The plaintiffs are not seeking damages but an emergency injunction: suspending the training and deployment of next-generation models until the defendants can prove their technology is under human control.
In response to the lawsuit, a Meta spokesperson said the company had no comment, and the other companies involved had not immediately responded to media requests for comment as of Reuters' publication.
This is not the first time the United States has sought through the courts to restrict AI model development. In June this year, Florida Attorney General James Uthmeier sued OpenAI, alleging safety problems with the ChatGPT platform, saying it provided information to individuals connected to a school shooting, offered self-harm guidance, and could make young users dependent.
In September this year, Florida further asked the judge overseeing the case to bar OpenAI from developing new AI models without external oversight. According to court records cited by Reuters, that request is still awaiting a ruling.
Regarding the Florida lawsuit, OpenAI spokesperson Drew Pusateri previously said the company had paused training of its most capable model and would not resume until additional safety safeguards were added. He also said the government should play an important role in setting AI safety standards.
In addition, in October this year, eight other U.S. states sued OpenAI in various courts, alleging that ChatGPT had guided users, including minors, to harm themselves or others. Those lawsuits seek court-imposed fines and restrictions on OpenAI's development of self-improving AI models unless the company can ensure those models align with human values.
Related articles
- Anthropic, OpenAI face patent lawsuit over alleged infringement of five neural network technology patents
- TSMC bets on AI: approves investment plan of up to $1 billion in Anthropic, OpenAI
- Former Anthropic Claude Code product lead Cat Wu jumps to OpenAI to oversee core product management
- OpenAI, Google, Microsoft and Anthropic jointly roll out AI data scraping opt-out standard
- "Largest wealth donation in modern history": Anthropic and two other AI giants plan to put nearly $300 billion into philanthropy
- Ahead of Anthropic's listing: report says OpenAI is accelerating a new funding round, with valuation possibly exceeding $1.4 trillion