In brief: Four paying users have sued Anthropic, OpenAI, Google and SpaceXAI over an alleged agreement to slow AI development. The filing turns a safety debate into an antitrust dispute. Public support on the same day does not by itself prove an agreement.
What the plaintiffs allege
According to the Associated Press, the complaint filed Friday in the US District Court for the Northern District of California seeks to represent a nationwide class against the four providers. The named plaintiffs each pay for ChatGPT, Claude, Grok or Gemini. They claim coordinated pacing reduces subscription value and violates Section 1 of the Sherman Act.
The complaint focuses on September 12. Anthropic CEO Dario Amodei published “We Must Pace the Frontier”, followed by public expressions of support from Sam Altman, Elon Musk and Demis Hassabis. The four companies had not responded to AP by Saturday. No court has tested or accepted the allegation.
The primary source is narrower than the allegation
Amodei's essay does call for industry coordination: shared safety standards and limits on unchecked progress. It also says regulation covering all US frontier labs would be more effective. For voluntary discussions, he asks for government mediation or a narrow antitrust waiver. He separately describes Anthropic's embedded external-evaluator pledge as unilateral.
Section 1 prohibits contracts, combinations or conspiracies that restrain trade. Several executives publicly endorsing the same position does not by itself establish a contract or private coordination. Yet Amodei's own request for a waiver shows that coordinated pace limits are legally different from routine safety-information sharing.
Pandorex Analysis
The new development is not the slowdown proposal – Pandorex has already analysed Amodei's three-stage plan – but the attempt to translate it into consumer harm. The plaintiffs still need to establish an actual agreement and attributable injury; simultaneous posts initially show timing, not causation.
The dispute exposes an institutional fault line. Unilateral safety reviews and generally applicable regulation avoid many antitrust risks. Voluntary capability or pacing limits among dominant rivals need clear government guardrails. At the editorial cutoff on September 20 at 18:18 Berlin time, the complaint itself was not yet available in the public RECAP/CourtListener index; claims about motions, a docket number or prospects would therefore be premature.