Antitrust law usually polices companies for competing too aggressively — price-fixing, market allocation, monopolization. A new class action filed in September 2026 alleges something almost the opposite: that four of the most powerful AI companies in the world illegally agreed not to compete as hard as they could.
What the lawsuit actually claims
The case, Buist v. Anthropic, PBC, was filed in the U.S. District Court for the Northern District of California and names Anthropic, OpenAI, Google, and SpaceXAI as defendants (case tracker, Tech Policy Law; docket monitor, Trace.law). According to reporting on the complaint, the plaintiffs — consumers who subscribe to these companies’ AI products — allege the companies reached an informal understanding to “pace” the release of more capable AI models rather than racing to deploy their most advanced systems as quickly as possible, and that this coordination was disclosed or evidenced through public statements and industry conduct rather than a written contract (CNN Business, Sept. 19, 2026; Bloomberg Law).
The legal theory rests on Section 1 of the Sherman Antitrust Act, the foundational federal statute that makes “every contract, combination… or conspiracy, in restraint of trade or commerce” illegal (15 U.S.C. § 1, Cornell Law School Legal Information Institute). Section 1 claims don’t require proof of a signed agreement — courts have long recognized that a “conscious commitment to a common scheme,” inferred from parallel conduct plus evidence of communication or motive, can be enough to get past early motions to dismiss. The U.S. Department of Justice’s own Antitrust Division explains the statute’s reach and enforcement approach in its Justice Manual (DOJ Justice Manual, Antitrust Statutes).
Why “agreeing to go slower” would even be illegal
This is the part that surprises non-lawyers: antitrust law doesn’t just punish agreements to raise prices. An agreement among competitors to limit output, restrain innovation, or otherwise avoid competing on a key dimension of their product — even one framed publicly as being about “safety” — can still be a per se or rule-of-reason violation if it wasn’t reached independently by each company for its own business reasons. The plaintiffs’ theory is that consumers were harmed by paying subscription prices for products that would have been more capable, sooner, absent the alleged coordination. Analysts following the case have noted this is one of the first major tests of antitrust law applied specifically to frontier AI safety coordination, an area regulators and courts have not yet had to address at this scale (CASRAI explainer).
Why this matters even if you don’t use these products
Cases like this will likely shape how every major tech company approaches public statements about “responsible” self-regulation going forward. Companies routinely make joint public commitments — on safety standards, environmental practices, content moderation — often coordinated through industry associations. This lawsuit tests how much coordination among direct competitors is legally safe versus how much starts to look like the kind of agreement Section 1 was written to stop. It’s an early-stage complaint, not a finding of liability, and defendants will have the opportunity to move to dismiss and to contest the allegations; the case is worth tracking rather than treating as settled fact.
This post is for general informational purposes and is not legal advice. This litigation is in its early stages and its allegations are, at this point, unproven — consult primary court filings or a licensed antitrust attorney for the current procedural status.
Sources
- Buist v. Anthropic, PBC — case tracker (Tech Policy Law)
- Buist v. Anthropic, PBC — docket monitor (Trace.law)
- CNN Business: AI slowdown antitrust lawsuit (Sept. 19, 2026)
- Bloomberg Law: OpenAI, Anthropic, Google, SpaceXAI hit with antitrust lawsuit
- 15 U.S.C. § 1 — Sherman Antitrust Act (Cornell Law School LII, official statute text)
- U.S. DOJ Justice Manual — Antitrust Statutes (official federal guidance)
