Sep 18 N.D. Cal. class action (3:26-cv-10693) alleges Anthropic, OpenAI, SpaceXAI, and Google agreed to slow AI improvement after Amodei’s pacing essay and public CEO responses. Plaintiffs seek treble damages for reduced subscription value. Allegation ≠ finding.

ABC7 News Bay Area — Lawsuit says Anthropic, OpenAI, SpaceXAI and Google made illegal agreement on AI slowdown · ABC7

What the video shows

Embed: ABC7 News Bay Area (YouTube AYMMYYOlqS8). Local TV / wire packaging of the complaint’s core claim. Prefer with hedges over NONE; send readers to the CourtListener PDF for caption, case number, and prayer for relief. A chyron that says “illegal agreement” is still describing a lawsuit theory.

What is new

AISN already covered Amodei’s pacing essay as a policy/opinion story. What is new is the docket: private plaintiffs converting public CEO alignment language into a consumer class action about subscription value. The scarce question is legal process — not whether AISN thinks slowing is wise.

What evidence supports

Complaint (primary, filed Sep 18, 2026). Captioned Charles Buist, Cheyenne Hunt, Christine Bullock, and Nick Spetsas v. Anthropic, PBC; OpenAI OpCo, LLC; SpaceXAI LLC; and Google LLC, Case No. 3:26-cv-10693, N.D. Cal. San Francisco Division. It alleges an agreement among frontier competitors to slow product improvement, cites Amodei’s September 12 essay and same-day public responses, and argues slower improvement is an output/quality restraint under Sherman Act §1. Plaintiffs say they do not challenge unilateral safety decisions — only alleged collective restraint. Remedies sought include treble damages (Clayton §4) and injunctive relief (Clayton §16) for a proposed nationwide class of paid individual subscribers to the four services.

Secondary coverage (Fortune, AP/OPB, ABC7). Confirm the filing date, defendant set, subscription-injury framing, and that company representatives had not immediately commented to those desks. Political pushback (Trump “hoax” framing; Hawley skepticism of antitrust exemptions) is backdrop, not proof either way on the merits.

What it does not prove

  • Not a cartel verdict or DOJ finding.
  • Not proof that public posts equal a completed illegal private agreement.
  • Not AISN’s judgment that pacing is good or bad policy.
  • Not SpaceXAI / Grok Bot product news — SpaceXAI is a named defendant in a consumer suit.

Why this matters

Public safety coordination talk and antitrust risk were already adjacent in Amodei’s own essay (waiver language) and Altman’s reply (proceeding without waiting for an exemption). A damages complaint makes that adjacency concrete for paid subscribers. Practical readers should track the docket — answers, motions to dismiss, any narrowing of “agreement” — not social-media scorekeeping about who “won” the slowdown argument.

What happens next

  1. Defendant answers or Rule 12 motions and any early case-management orders.
  2. Whether DOJ/FTC take any parallel public interest — separate from this private suit.
  3. How labs describe coordination vs. unilateral safety work in future statements.

Bottom Line

Buist (3:26-cv-10693, filed Sep 18) turns “pace the frontier” talk into an alleged Sherman Act coordination case about subscription value. A filed complaint is a real legal event. It is not a proven cartel verdict. Keep allegation and adjudication in different columns.

Sources