Consumers sue Anthropic, OpenAI, SpaceXAI, and Google alleging antitrust pact to slow AI development

Four subscribers to paid AI assistants filed a proposed class action in the Northern District of California on September 18, 2026, alleging that Anthropic, OpenAI, SpaceXAI and Google agreed among themselves to slow the pace at which each improves the products it sells to consumers. The complaint, Buist v. Anthropic, PBC, No. 3:26-cv-10693, is Document 1 on the docket and runs 29 pages; it demands a jury. The named defendants are Anthropic, PBC; OpenAI OpCo, LLC; SpaceXAI LLC; and Google LLC.

The plaintiffs are Charles Buist and Nick Spetsas of Florida and Cheyenne Hunt and Christine Bullock of California. Buist, Hunt and Spetsas each allege they bought paid individual subscriptions to Claude, ChatGPT, Grok and Gemini during the class period; Bullock alleges a subscription to Claude (¶¶ 16 to 19). The complaint is signed by Andrew T. Tutt of Trial Lawyers for Justice, with Nicholas C. Rowley listed as lead counsel; two further attorneys, R. Stanton Jones and Jakob Z. Norman, are marked as having pro hac vice applications forthcoming.

The claims and the relief sought

There are two claims for relief. The first alleges a violation of Section 1 of the Sherman Act, 15 U.S.C. § 1, against all defendants (¶¶ 135 to 145). The second seeks injunctive relief under Section 16 of the Clayton Act, 15 U.S.C. § 26 (¶¶ 146 to 152).

The first claim is pleaded three ways. Paragraph 140 calls the agreement “a naked horizontal restraint on output and product quality among competitors,” not ancillary to any integration or joint venture, and “therefore unlawful per se.” Paragraph 141 pleads, in the alternative, that it is unlawful under “quick-look” analysis because “an observer with even a rudimentary understanding of economics would conclude” that an agreement among a market’s principal competitors to slow technological progress harms consumers. Paragraph 142 pleads the rule of reason, alleging that the defendants “possess substantial collective power,” that the agreement has “no cognizable procompetitive justification,” and that any legitimate objective could be met by less restrictive means, including unilateral safety decisions and independent evaluation.

The prayer asks the court to certify the class and subclasses; declare that the defendants joined an unlawful contract, combination or conspiracy under Section 1; award damages “in an amount to be established at trial,” trebled under Section 4 of the Clayton Act, 15 U.S.C. § 15; enter preliminary and permanent injunctions; and award interest, fees and costs. Paragraph 150 lists what the injunction would prohibit: any agreement with a horizontal competitor concerning the rate at which competing AI products are developed or released; limits on training compute or training runs adopted under a pacing arrangement; limits on using AI to develop improved AI where imposed by agreement among competitors; coordinated release delays; capability checkpoints that operate as agreed restraints; and exchanges of competitively sensitive information to police them.

The market and the class

“To the extent market definition is required,” the complaint defines the product market as “paid consumer subscriptions to general-purpose frontier generative-AI assistants,” which it calls the “Paid Frontier AI Assistant Subscription Market,” naming the paid tiers of ChatGPT, Claude, Grok and Gemini as its principal products (¶ 96). The geographic market is the United States (¶ 101). On information and belief, the defendants “collectively account for at least 80%” of that market (¶ 102). Paragraph 103 lists compute, specialized semiconductors, data-center capacity and talent among the barriers to entry. The only price the complaint states is that ChatGPT Plus is “currently approximately $20 per month” (¶ 26).

The proposed nationwide class, under Rules 23(a), 23(b)(2) and 23(b)(3), is “All persons in the United States who, from September 12, 2026 through the date on which the effects of Defendants’ unlawful conduct cease,” bought directly from a defendant “a paid individual consumer subscription providing premium access to ChatGPT, Claude, Grok, or Gemini” (¶ 123). Paragraph 124 proposes one subclass of direct subscribers per defendant.

The alleged agreement

The class period begins on September 12, 2026, the morning Anthropic’s chief executive Dario Amodei published an essay titled We Must Pace the Frontier (¶¶ 2, 61). It quotes the thesis, “We must slow the pace at which we improve the capabilities of AI models” (¶ 62), and describes three proposed steps. The first, embedding third-party evaluators with employee-like access to models and training pipelines, Anthropic adopted on its own (¶ 63). The second, which the essay called “Democratic Coordination,” “requires industry-wide coordination” to set common safety standards and “limits on the rate of unchecked AI progress” (¶ 64). The complaint says the essay proposed limits on training compute, on the nature of training runs and on the internal use of AI to improve AI, and capability “checkpoints” (¶ 65); that embedded evaluators would make “verifiable pacing” practical (¶ 66); that coordinated pacing would let developers slow “without sacrificing commercial advantage” (¶ 67); and that the essay urged companies to “voluntarily work together” in parallel with any government action (¶ 68). Paragraph 69 alleges the essay acknowledged that the government would need to issue a narrow waiver for certain safety conversations, and states: “No such waiver exists.”

The complaint alleges that the other three companies accepted the proposal in public the same day. Within about an hour, Elon Musk, whom the complaint describes as the person who founded and controls SpaceXAI’s AI business, quote-posted the announcement with “Dario is right” (¶¶ 71, 85). OpenAI’s chief executive Sam Altman wrote, “I agree with Dario that we need to pace the frontier,” and said OpenAI “will do the same” in giving independent evaluators employee-like access (¶ 72). Demis Hassabis of Google DeepMind said the essay pointed toward “the right path forward” and that “the direction is correct,” tying it to his proposed “industry-wide standards body for frontier AI” (¶ 73).

The complaint also pleads earlier conduct. It alleges a July 2026 statement titled Pacing the Frontier, signed by Amodei, Anthropic co-founder Jared Kaplan, OpenAI chief scientist Jakub Pachocki, OpenAI chief research officer Mark Chen and Google DeepMind co-founder Shane Legg, among others, which described “intense competitive pressure not to unilaterally slow” (¶¶ 49, 50); a July 14 proposal by Hassabis for a U.S.-led standards body modeled in part on FINRA (¶ 53); a working group of Anthropic, OpenAI and Google representatives below the chief executive level that met regularly from July, which The Information reported on September 13 was still meeting (¶ 55); a September 6 essay by Pachocki naming “coordinating to slow down future development” as an option for frontier developers (¶ 57); a September 10 WIRED report that OpenAI had asked members of Congress whether coordinating a slowdown could violate the antitrust laws (¶ 58); and Altman’s answer in a September 11 Fortune interview, when asked why AI executives did not develop a common plan: “I think that will happen” (¶ 59).

After September 12, the complaint alleges, Altman wrote on September 14 that AI progress “should be slower than it otherwise could be” and that OpenAI would not wait for an antitrust exemption or legislation (¶¶ 77, 78), and OpenAI’s global policy chief Chris Lehane confirmed on September 15 that OpenAI had worked with Anthropic and Google DeepMind on these issues for several weeks before the essay (¶ 79).

What the complaint says about the defenses

On safety, the complaint states that the plaintiffs do not challenge any defendant’s unilateral decisions about safety, testing or the pace of its own development (¶ 13), that each defendant can do all of those things unilaterally (¶ 111), and that “The Act does not permit competitors to decide among themselves that competition is undesirable as a policy matter; that judgment belongs to Congress” (¶ 112). On petitioning, the plaintiffs “do not seek to impose liability for any Defendant’s advocacy to Congress, the White House, or any agency” (¶ 88); the restraint challenged is “Defendants’ private agreement among themselves, formed and implemented outside any governmental process” (¶ 89); and petitioning statements are alleged only as evidence of knowledge and intent (¶ 90). On exemption, it alleges that no statute exempts the agreement and no governmental body compelled it (¶¶ 69, 114).

The injury alleged is delay in improved capabilities, reduced product quality and reduced investment (¶ 107), and subscription prices “that exceed the competitive value of the products they receive,” an overcharge the complaint says is recoverable under Section 4 of the Clayton Act (¶ 117). It states that the amount of damages “will be established through discovery and expert proof” (¶ 118).

Sources and reporting

The facts in this piece are taken from the complaint as filed; paragraph numbers refer to it. Statements by Amodei, Musk, Altman, Hassabis, Pachocki and Lehane, the July statement and the WIRED, Fortune and The Information reports are given as the complaint quotes or describes them; those underlying documents were not independently reviewed. Bloomberg Law, Politico and The Hill reported the filing and do not conflict with it. Bloomberg Law reported that none of the defendants immediately responded to a request for comment.

What is not yet known

No defendant had answered, moved against the complaint or commented on it publicly in the sources reviewed. The complaint’s caption names no assigned judge, and none of the reporting identifies one. The complaint does not allege any written agreement among the defendants; it pleads formation from public statements, the July statement and the working group. It does not plead an amount of damages. The pro hac vice applications for two of the plaintiffs’ attorneys had not been filed as of the complaint’s date, September 18, 2026.