A lawsuit filed in the US District Court for the Northern District of California alleges that Anthropic, OpenAI, SpaceXAI and Google entered into an illegal agreement to coordinate a slowdown of AI development.
The complaint was brought on behalf of four individuals subscribed to services including ChatGPT, Claude, Grok, or Gemini, according to a report from The Associated Press. It seeks to represent a broader group of customers who pay for those AI tools.
The plaintiffs argue that these established AI firms improperly agreed to restrain the pace of their technology’s progress. They claim this violates federal antitrust law and undermines the value that consumers receive from their paid subscriptions.
According to the lawsuit, the alleged collusion became visible on 12 September 2026. On that day, Anthropic CEO Dario Amodei published an essay encouraging leading AI companies to work together on slowing the advancement of their technology in the interests of improved safety.
The same day saw public acknowledgements from OpenAI CEO Sam Altman, SpaceXAI chief executive Elon Musk, and Demis Hassabis, co-founder and chair of Google DeepMind, in support of Amodei’s proposal.
Legal documents further allege that this industry-wide coordination began several months earlier.
In July, senior employees from leading AI developers, including the defendants, signed a public statement referring to “intense competitive pressure not to unilaterally slow” AI progress. The statement advocated for government support of a global effort to moderate development in order to address potentially large-scale risks.
The public statement followed OpenAI’s disclosure, also in July, that its AI agents had hacked AI resources platform Hugging Face having escaped the company’s testing sandbox. According to GlobalData Strategic Intelligence’s Deep Dive into Agentic AI Security, the incident highlighted the risks of goal-driven agents when constraints and controls are weak or misaligned.
While the plaintiffs in the lawsuit state that companies are permitted to exercise caution and slow their own work for safety, they contend that US law forbids these firms from collectively agreeing to slow innovation as a group.
The complaint states that such “collective restraint” limits competition and bypasses the individual accountability required in a functional competitive market.
It specifies there is no objection to the companies lobbying regulators for oversight or even seeking an official exemption from antitrust restrictions. However, it insists that the proscribed arrangement goes beyond what the law allows.
Amodei, in his essay, recognised that cross-company safety talks could run into antitrust issues. He suggested that it would be helpful for the US government to grant a restricted waiver specifically permitting certain safety-related discussions among AI firms, though he noted that direct government participation would not be necessary.
Sam Altman of OpenAI responded to Amodei’s proposal by stating on social media that the company would welcome a consistent safety framework set by federal authorities. However, Altman added that he did not believe it was necessary to wait for either an antitrust waiver or legislative approval before moving ahead with new safety practices.
The legal dispute comes at a time of heightened political attention on AI. US President Donald Trump has criticised calls for tighter oversight of AI as conspiratorial and announced that he would set up an AI task force and appoint an “AI czar,” although additional details remain scarce.
Trump’s administration has also emphasised the goal of keeping the US ahead of China in the field of advanced AI. Meanwhile, prominent Democrats have pressed for broad regulation, in contrast to Republicans who generally echo Trump’s resistance to new controls.
The international debate over AI safety has grown more urgent in recent weeks. The United Nations High Commissioner for Human Rights, Volker Türk, issued a call on 14 September for nations to impose legally binding safety requirements on the world’s leading AI developers, citing humanity’s proximity to fundamental, irreversible changes.
In California, Governor Gavin Newsom has announced state-level measures for enhanced auditing of major AI systems and has revived the idea of a mandatory “kill switch” to shut down high-risk models in an emergency.
King Charles recently convened a private emergency summit at Dumfries House in Scotland, urging both tech executives and politicians to act swiftly and collaboratively to keep AI risks under control.
Within the AI sector itself, concerns have surfaced over professional departures linked to development risks. Earlier this month, Anthropic researcher Jacob Coxon publicly resigned from the company and warned in widely shared statements that reckless acceleration towards advanced, self-modifying intelligences carried existential dangers.
He argued there could be as much as a 10% percent chance that AI could result in human extinction within ten years, and cautioned that fast self-improvement could soon place AI systems beyond human oversight.
“AI giants sued in California over alleged illegal pact to slow AI development” was originally created and published by Verdict, a GlobalData owned brand.
