Lawsuits challenging how major U.S. AI companies develop their technology are arriving from opposite directions. On November 12, 2026, four nonprofit organizations sued Anthropic, Google, Meta, OpenAI and SpaceXSI in a California court, seeking to block the development and deployment of new AI models. Earlier, on September 18, four paying users of AI services filed an antitrust suit in federal district court, alleging that Anthropic, OpenAI, Google and SpaceXAI had coordinated to slow AI development in violation of antitrust law. One suit warns of the dangers of continuing to build AI; the other objects to competitors jointly holding development back. Both reflect only the plaintiffs' claims, and no court has found any illegality, but together they highlight how hard it is for AI companies to balance safety and competition.
In September and November, three AI giants were sued from opposite directions
The November 12 suit was filed by four groups: Irreplaceable Education Fund, Public Citizen, Heat Initiative and Alliance for Secure AI. According to an announcement from Public Citizen, one of the plaintiffs, the case was filed in California Superior Court in San Francisco County. The plaintiffs argue that the major AI companies have continued to develop and offer their technology while aware of its serious dangers, and they are seeking an injunction under California's unfair competition law.
According to AP reporting, the plaintiff groups contend that cyberattacks and other harms enabled by advanced AI systems threaten public health and interfere with the use of property, including computer systems. The complaint also points to the fact that AI company executives themselves have warned about the technology's potential dangers. The plaintiffs argue that developing products capable of causing serious harm should not be treated as inevitable.
Matthew Platkin, a former New Jersey attorney general representing the plaintiffs, criticized the companies for pressing ahead despite recognizing catastrophic risks, saying they continue conduct that endangers the public. Al Jazeera reports that the plaintiffs are seeking an emergency injunction to immediately halt development and deployment of a new generation of AI models. They reportedly want development and deployment to stay suspended until the companies can prove the AI systems remain under human control.
The suit filed on September 18, by contrast, takes aim at coordination among companies to slow AI development. According to Bloomberg Law, four paying subscribers of ChatGPT, Claude, Grok and Gemini sued Anthropic, OpenAI, Google and SpaceXAI, alleging violations of Section 1 of the Sherman Antitrust Act. The plaintiffs argue that if competitors jointly hold back the pace of development, competition weakens and improvements to the products users pay for are delayed.
The suit was triggered by an essay by Anthropic CEO Dario Amodei, published on September 12, calling for a slowdown in AI development. That same day, OpenAI's Sam Altman, SpaceXAI's Elon Musk and Google DeepMind's Demis Hassabis voiced support for the proposal, which the plaintiffs cite as evidence of coordination among the companies. However, executives publicly endorsing the same idea is a separate matter from an agreement restricting competition actually being reached. The plaintiffs also do not object to companies slowing development on their own; what they challenge is a joint agreement among competitors to hold development back.
In the 55 days between September 18 and November 12, Anthropic, OpenAI and Google became defendants both in a suit that calls coordinated slowdowns illegal and in one that calls continued development illegal. SpaceX-related entities appear in both suits, but they are referred to as SpaceXAI in the September coverage and SpaceXSI in the November coverage, and it has not been confirmed whether they are the same legal entity, so they are not counted among these three companies. Meta is a defendant only in the November suit. No court has ruled on the legality of the conduct in either case, and the companies' development activities have not currently been prohibited.
AI companies' options: slow down on their own, or follow common rules
Considering the questions raised by the two suits together, there are broadly two ways AI companies could adjust the pace of development. One is for each company to assess safety independently and delay development or release as needed. The other is to impose common constraints, separate from voluntary agreements among companies, through laws, regulations or court orders. This is a framing drawn from the claims in the two suits, however, and the legal options are not strictly limited to these two.
The September antitrust suit does not take issue with individual companies putting safety first on their own. A company that slows development based on its own safety evaluation, without any agreement with competitors, can therefore at least avoid the inter-company coordination that suit challenges. The November suit, on the other hand, seeks to block the development and deployment of dangerous new models altogether, so simply slowing the pace of development might not satisfy the plaintiffs' demands. If the requested injunction were granted, the affected models would have to stay suspended until the conditions set by the court were met.
OpenAI offers an example of a company halting development on its own. When Florida asked a court at the end of September to bar OpenAI from developing new models without outside oversight, OpenAI spokesperson Drew Pusateri said the company had stopped training its most capable model on September 25. According to Bloomberg Law, OpenAI had indicated it would not resume training until it was confident additional safeguards were in place.
OpenAI has also shelved the release of its next-generation model, "GPT-6.1 Astra." According to a Reuters report on September 28, problems were identified in internal safety evaluations, and the release planned for October was called off. These are cases of individual companies halting development or release for safety reasons, but it has not been confirmed whether the training halt announced in September is still in effect. Nor is it certain that such voluntary measures alone would satisfy the conditions the November suit seeks.
The other approach is the introduction of common rules by government or the courts. Phil Aroneanu, founder and executive director of Irreplaceable Education Fund, told Al Jazeera that an effective halt to development would require clarity on the scope of the work being prohibited, how compliance would be verified, and what the consequences of violations would be. That, he said, would require government inspections and the authority to stop prohibited work.
If common safety regulations were put in place, companies could be required to meet certain safety standards even without voluntarily agreeing among themselves to slow down. But difficult questions remain: how far regulation should extend across training, release and operation of models, how compliance would be verified, and what conditions would end a suspension. A court ordering an injunction against individual companies and a legislature building a system that applies across the industry would also differ in their scope and legal effect. To make a halt on development effective, it would not be enough to simply order a stop; a concrete oversight structure would have to be designed.
Political barriers to AI regulation and what to watch next
There are also political challenges to establishing rules that apply across the industry. According to the AP, President Trump has supported rapid AI development, emphasizing that the U.S. should stay ahead of China and others. While he has favored voluntary efforts by companies on safety measures, he has also shown some understanding of additional regulation. Congress has not yet passed comprehensive AI regulation, but with Democrats having regained control of Congress in the 2026 midterm elections, discussions on regulation could advance in the new Congress that convenes in 2027.
The U.S.-China race in AI development also complicates regulation. Irreplaceable Education Fund argues that if the U.S. and China slowed development to a comparable degree, risks could be reduced without undermining either country's relative competitiveness. It says achieving this would require a bilateral agreement or international treaty. However, there is no guarantee that both countries would accept regulation on the same terms, and no concrete path to such an international agreement has been laid out.
Meanwhile, a company that slows development on its own may face competitive disadvantages. The September antitrust suit also cited a statement signed in July by executives of AI labs. The statement pointed to intense competitive pressure that makes it hard for any one company to slow down alone, out of fear of being overtaken by rivals. Regardless of how the lawsuits turn out, the possibility remains that a company that halts development for safety reasons could lose market share or technological advantage.
Still, the two suits do not mean AI companies must now immediately choose between halting development and accepting common regulation. In the September case, the plaintiffs must prove that an unlawful agreement existed among the companies, and in the November case, a court will examine whether there is sufficient legal basis and need for the injunction the plaintiffs seek. If either claim is dismissed, the picture of two lawsuits simultaneously constraining company behavior would weaken. Courts could also reach different decisions depending on each company's safety measures and development conditions.
The key milestones ahead are whether the court grants an emergency injunction in the November case, and whether unlawful coordination among the companies is found in the September antitrust case. Also worth watching is whether the new Congress begins deliberating legislation to set common safety standards and an oversight system for AI companies. The courts' decisions and congressional action should clarify whether each company will continue with its own safety measures or follow common standards set by the government.
The suits are not irrelevant to users in Japan either. The defendants include the developers of AI services widely used in Japan, such as ChatGPT, Claude and Gemini. If an injunction against development or deployment were granted, depending on the scope of the order, it could affect development plans and release timing for new models. However, it has not been decided that existing services would be suspended, and whether an injunction is granted and how far it extends will be left to future court rulings.
The two lawsuits show that the questions of safety and competition in AI development are increasingly too big to be settled as matters of corporate management alone. Coordination among companies may raise antitrust concerns, while companies that continue development independently face questions about their legal responsibility for safety. Going forward, not just the outcomes of the individual cases but whether a common regulatory framework emerges that lets companies compete while ensuring safety will be a key issue shaping the direction of the AI industry as a whole.