Lawsuit Accuses Anthropic, OpenAI, SpaceXAI and Google of Illegal AI Slowdown Agreement

Key Points

1. Four subscribers allege rival AI companies coordinated to slow development, harming competition and paid users.
2. Plaintiffs accept independent safety slowdowns but argue collective restraint by competitors violates US antitrust law.
3. The case could shape how AI developers cooperate on safety while competing for users, talent and investment.

The latest

Anthropic, OpenAI, SpaceXAI and Google face a proposed nationwide antitrust class action filed in the US District Court for the Northern District of California. Four paying subscribers to ChatGPT, Claude, Grok and Gemini accuse the companies of coordinating efforts to moderate AI development, limiting competition and reducing the value of their subscriptions. The plaintiffs seek to represent paying users across the United States.

Details

• Legal boundary: The plaintiffs do not contest each company’s right to delay its own systems for added testing or safeguards. Their case draws a line between unilateral safety decisions and an alleged agreement among competitors to reduce the pace of innovation collectively.
• September exchanges: The complaint focuses on Anthropic CEO Dario Amodei’s public call in September for deeper industry cooperation to slow some advances while reinforcing safety. OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Google DeepMind co-founder and chairman Demis Hassabis later made public statements broadly supporting greater safety coordination.
• Earlier statement: In July, senior employees from several leading AI laboratories signed a statement describing intense competitive pressure that prevented individual companies from slowing development alone. It sought government backing for an international effort to moderate automated AI development, an episode the lawsuit portrays as part of an effort already under way months earlier.
• Safety rationale: Amodei has argued that safety cooperation is needed because a laboratory acting alone could lose market share, talent and investment as rivals continue advancing. The plaintiffs recast that pressure as an antitrust concern, alleging competitors could remove it through a shared slowdown rather than separate risk decisions.
• Regulatory options: Amodei acknowledged possible legal friction, proposing government mediation or a narrow antitrust exemption for specific safety discussions. Altman supported a federal framework with consistent safety requirements but argued that companies should not wait for legislation or an exemption before cooperating on safety.
• Policy pressure: Washington remains divided over AI controls. The Trump administration has prioritised US leadership against China and opposed broad restrictions that could slow American companies. Trump has criticised AI guardrails while announcing plans for an AI task force and an AI czar; lawmakers in both parties have questioned coordination outside a clear framework.
• Market stakes: ChatGPT, Claude, Gemini and Grok compete for millions of paying users while their developers invest billions in chips, data centres, computing infrastructure and model training. An alleged coordinated slowdown could influence product launches, investment choices, computing demand and competition across the market.

What’s next

The court’s decision on whether the four subscribers may proceed on behalf of a nationwide class will be the next concrete procedural marker, alongside its assessment of where safety cooperation crosses into unlawful collective restraint.