Buist v. Anthropic: Docket Advances — Summonses Issued, Google Cast as Architect of the "Pacing" Coordination Venue

Buist v. Anthropic: Docket Advances — Summonses Issued, Google Cast as Architect of the "Pacing" Coordination Venue

Buist v. Anthropic PBC, OpenAI OPCO, SpaceXAI LLC, Google LLC (No. 3:26-cv-10693, N.D. Cal.) — the Sherman Act §1 class action over the alleged industry agreement to slow AI development — has moved into active litigation. Per FinanceFeeds' September 22 reading of the full 29-page complaint and the CourtListener docket: summonses were issued September 21, 2026, and the case sits before Magistrate Judge Nathanael M. Cousins. No defendant had responded formally as of that writing; no motion to dismiss has yet been filed.

New detail on Google's alleged role — the complaint's paragraph 84 alleges Google "took part in a working group from July, proposed the standards body through Hassabis, had DeepMind co-founder Shane Legg sign a July statement conceding 'intense competitive pressure not to unilaterally slow,' and on 12 September 'endorsed the direction' of Amodei's proposal 'while directing its implementation toward the body Google had proposed.'" The plaintiffs' theory targets the venue, not just the pact: Hassabis's proposed FINRA-style standards body has no statutory basis, and under Silver v. New York Stock Exchange (1963) even Congressionally-authorized self-regulation escapes the antitrust laws "only to the minimum extent necessary." The complaint's summary of the alleged agreement: it was "proposed in public, accepted in public, and confirmed in public."

Relief sought: treble damages under Clayton Act §4 for a nationwide class of paid ChatGPT, Claude, Grok and Gemini subscribers from September 12, 2026 onward (defendants pleaded at "at least 80%" of U.S. paid frontier AI assistant subscriptions), plus an injunction barring any horizontal agreement on "limits on training compute," timing of training runs, or "capability checkpoints" — while carving out unilateral slowing, independent evaluators, and lawful petitioning of government (¶¶150–151).

The defense evidentiary problem and its limit: the filing "contains 'no internal documents, no confidential witnesses, and no meeting minutes'" (The Next Web) — the record is the defendants' own public statements. Altman's most exposed line: "We do not believe we need to wait for an antitrust exemption or legislation to begin the work of providing this confidence." Against that, Washington has offered no exit: President Trump dismissed calls for regulation while forming an AI task force, and Senator Hawley said "there is no world" in which he would grant an antitrust exemption. The complaint itself invokes National Society of Professional Engineers (1978): the judgment that competition is unsafe "belongs to Congress."

Watch posture: this remains an early-stage case — the pivotal next events are the defendants' motions to dismiss (expected to turn on whether endorsing a "direction" plus proposing a standards body constitutes an agreement on a "rate of development") and any sign of the alleged cross-lab working group's continued activity, which would feed the plaintiffs' record. The case's structural significance for enterprises: safety coordination between AI vendors is now litigated as an output restraint, which cuts both ways for deployers — it forecloses industry-wide pacing pacts that might stabilize model risk, while leaving unilateral vendor caution untouched.