Anthropic, OpenAI, SpaceXAI and Google are facing a new antitrust lawsuit alleging that the companies coordinated to slow the pace of AI development. The complaint argues that weaker-than-normal competitive pressure could reduce the value of paid services including ChatGPT, Claude, Grok and Gemini.
The case was filed on September 18 in the U.S. District Court for the Northern District of California by four users who subscribe to the services named in the complaint. They are seeking to represent other paying users across the country in a proposed class action. The court has not yet decided whether the case may proceed on a class-wide basis.
Public comments from four companies underpin the case
The complaint identifies September 12 as a key date in the alleged coordination. On that day, Anthropic CEO Dario Amodei published an article calling on the AI industry to slow technological progress collectively and devote more resources to safety measures. OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Demis Hassabis, co-founder and chairman of Google DeepMind, later responded publicly and expressed support for the direction Amodei outlined.
The plaintiffs argue that several major competitors publicly endorsing an AI development pace below the level that would emerge from normal market competition could amount to coordinated restraint. They say such conduct could affect the speed of model updates, improvements in product capabilities and the value users receive after paying for subscriptions.
Nick Rowley, the lead attorney representing the plaintiffs, said that if AI safety standards and related agreements are controlled through private arrangements among a small number of profit-driven technology companies, consumers could face consequences they cannot readily assess. He also linked the risk of AI systems escaping human control to coordination between companies, although those statements remain allegations made in the lawsuit.
As of September 19, Anthropic, OpenAI, Google and SpaceXAI had not immediately responded to the requests for comment. The court had also made no ruling on the merits of the allegations.
Amodei raised the possibility of government involvement
In the article calling for slower development, Amodei acknowledged that such cooperation could raise antitrust concerns. He said the U.S. government could coordinate safety discussions between different AI laboratories, or at least facilitate certain types of safety-related communication. The government would not need to participate in every discussion, he argued, but could provide an exemption for narrowly defined safety issues.
Altman later said OpenAI supported a federal framework establishing consistent safety requirements. He added that the company did not believe it needed to wait for an antitrust exemption or new legislation before taking steps to build public confidence.
The dispute comes as AI companies and researchers increasingly examine whether advanced models could operate beyond human control. Discussions about common safety standards, coordinated testing methods and the sharing of safety experience did not begin in September. But when statements from competing companies are viewed together on the same timeline, the boundary between legitimate cooperation and anticompetitive coordination becomes relevant to both paying users and regulators.
Plaintiffs say regulation itself is not the issue
The plaintiffs stress that they do not oppose Anthropic, OpenAI, SpaceXAI or Google making AI policy recommendations to Congress, the White House or other government bodies. They also do not object to companies seeking an antitrust exemption. Their challenge is whether companies may coordinate among themselves to slow product and technological competition without a clearly defined government framework.
The U.S. government remains divided over the appropriate approach to AI regulation. At the federal level, policymakers have promoted efforts to keep U.S. AI laboratories technologically competitive, while some lawmakers have warned that the industry's most influential companies should not use safety cooperation to obtain antitrust protection and collectively restrict competition. Policy discussions are continuing, and no definitive framework has been announced for setting the boundaries of cross-company safety discussions.
For the market, the case concerns more than whether companies can discuss AI safety. It also raises questions about the pace of product updates and the competitive structure of paid AI services. If slower development ultimately produces fewer improvements in model capabilities, subscribers may question whether prices remain aligned with the services they receive. At the same time, an outright ban on sharing safety information could create additional risks during model deployment if companies genuinely need to exchange that information. Those issues will need to be assessed alongside corporate filings, actual product changes and subsequent court proceedings; the allegations in the lawsuit have not been judicially established.