Subscribers Sue the AI Big Four: A Coordinated Slowdown Just Became an Antitrust Case

The day the AI heavyweights shook hands "for safety" was the day they handed plaintiffs an antitrust complaint.

Three key facts

On September 18, a group of paying subscribers filed suit in the U.S. District Court for the Northern District of California against Anthropic, OpenAI, SpaceXAI and Google, alleging the four violated antitrust law by "coordinating to slow AI development." The plaintiffs subscribe to ChatGPT, Claude, Grok and Gemini; they argue the coordination restricts competition and dilutes the value their subscription fees buy, and they seek treble damages and injunctive relief.

The complaint anchors on one date — September 12: Anthropic CEO Dario Amodei published an essay that day calling for industry-wide cooperation to slow AI progress in exchange for stronger safety measures; OpenAI's Sam Altman, SpaceXAI's Elon Musk and Google DeepMind's Demis Hassabis all publicly agreed the same day. The legal logic fits in a sentence: one company choosing to slow down for safety is fine, but rivals agreeing together to slow down is textbook cartel behavior — the complaint calls it "substituting collective restraint for individual accountability."

The suit also digs up an earlier thread: in July 2026, senior employees across several AI labs signed a joint statement admitting "intense competitive pressure not to unilaterally slow" development, and calling for government support of a global deceleration effort. The market-structure number the plaintiffs cite: the four defendants collectively control at least 80% of the U.S. paid AI assistant subscription market.

Wang Dou's Take

The sharpest irony in this case: everything the defendants did was, viewed alone, a PR plus — caring about safety, urging restraint, industry self-discipline. Antitrust law has never cared about motives; it cares about coordination. Four competitors holding 80% of a market nodding at each other on the same day about "slowing down together" is called a horizontal agreement in court, not a gentlemen's accord. The safety narrative served as a get-out-of-jail card for three years; now someone has entered that very card as Exhibit A.

The angle the plaintiffs picked is also cunning: they don't sue over monopoly pricing — they sue over "shrunken subscription value." You pay $20 a month for a model that keeps getting better; when four rivals agree to get better slower together, that's a collective downgrade with no price cut. It translates the safety debate into consumer harm — a language juries actually understand.

For the industry, the real signal is that the "self-regulation" road just got half-blocked. Labs used to slow down whenever safety made a convenient excuse; now every deceleration carries the question of whether it becomes evidence. That leaves one respectable path: if you want to slow down, go push governments to write the rules — public rulemaking, clear accountability, still safer than four CEOs nodding across the internet. California is legislating, Congress is brawling, and the courts have started accepting filings: the battle over who referees AI safety officially began this week.

Source: Los Angeles Times · CBS News · Tom's Hardware