OpenAI, Anthropic, xAI and Google sued over pact to slow AI: coordinated caution collides with antitrust law

By Zendoric · September 27, 2026.

Four people who pay for subscriptions to ChatGPT, Claude, Grok or Gemini have sued OpenAI, Anthropic, Google and xAI —Elon Musk's artificial intelligence company, maker of Grok, which the outlet that broke the story, the North State Journal, refers to as "SpaceXAI"—. The lawsuit, filed last Friday in the U.S. District Court for the Northern District of California on behalf of a proposed nationwide class of subscribers, does not accuse these companies of building bad products. It accuses them of agreeing among themselves to move more slowly, something the complaint says violates antitrust law —the rules that bar competitors from coordinating instead of competing— because it reduces the value received by those who pay for those subscriptions.

The trigger the lawsuit points to is specific and dated. On September 12, Anthropic CEO Dario Amodei published an essay calling for industry-wide cooperation to deliberately slow the pace of advances in favor of greater safety. That same day, according to the complaint, Sam Altman (OpenAI), Elon Musk (xAI) and Demis Hassabis (co-founder and president of Google DeepMind) responded publicly voicing their agreement. For the plaintiffs, that exchange was not a coincidence of opinions but the visible proof of a coordination that had been taking shape since July 2026, when senior figures from several AI labs signed a joint statement acknowledging the "intense competitive pressure not to slow down unilaterally" and asking the government to back a global deceleration effort.

The legal nuance is what makes the case interesting. The plaintiffs are not saying these companies cannot slow their own development for safety reasons: they are saying they cannot do it together and by mutual agreement. The complaint argues that antitrust law prevents rivals from taking the "shortcut" of "substituting individual responsibility with a collective restraint"; a competitive market, they argue, is precisely what forces each company to answer for its own pace. "If we allow AI safety and protocol... to be controlled by private, self-interested agreements among the most powerful tech companies in the world, AI will quickly escape human control and could kill us all," said Nick Rowley, lead attorney for the plaintiffs. Neither Anthropic, OpenAI, Google nor xAI had responded to the outlet's requests for comment at the time of publication.

Amodei himself had anticipated the legal problem in his essay: he wrote that it would be helpful for the U.S. government to mediate —or at least enable— these conversations among rival labs, which would require only a "limited exemption" (a specific legal permission, known as a waiver, authorizing conduct that is normally prohibited) for certain safety talks, without Washington having to participate directly. Altman responded that OpenAI would welcome "a federal framework setting consistent safety requirements," but clarified that he does not think it necessary to "wait for an antitrust exemption or a law" to start building that trust. In other words: even the signatories of the alleged pact disagree on whether they need legal permission to coordinate.

That permission, for now, is not available and probably will not be anytime soon. President Donald Trump rejected the calls for regulation on social media, calling them a "conspiracy," and wondered why industry leaders would ask for rules that, in his view, if strictly enforced, would drive them "to ruin and bankruptcy." On Saturday he announced he will form an AI task force and appoint an "AI czar," without further details. From the president's own party, Republican Senator Josh Hawley was blunt at a recent Senate hearing: there is "no world" in which he would agree to exempt "the most powerful companies in history" from antitrust laws, precisely out of fear that they would collude and choke off competition. The plaintiffs, for their part, insist they do not oppose Congress or the White House legislating on AI, nor companies seeking that exemption through the proper channels.

It is worth separating the layers here: one is the real technical risk of an AI advancing faster than it can be audited; another is the market structure being built around that risk, with players who benefit from appearing prudent without necessarily slowing down at all. This lawsuit adds a third layer, the legal one, and it is the one that may prove most decisive in the short term. If coordinating publicly on safety can land you in court accused of collusion, the natural incentive for any prudent CEO is to stop doing it in public, or to stop doing it altogether, and return to the usual competitive logic. The lawsuit, filed to force greater individual accountability, risks producing the opposite effect to the one it seeks: pushing companies back into the "if I don't do it, my rival will" pressure that the July 2026 statement itself described as the underlying problem.

This is no minor matter for the underlying thesis with which we follow this sector. We believe AI can lead us to real abundance —health, more years of healthy life, freedom from routine work— but only if the transition there is managed with legitimate governance, not with backroom deals among a handful of for-profit companies. In that sense, the plaintiffs' instinct to demand individual, verifiable accountability rather than an opaque collective agreement is not the enemy of safety: it may be the only path that survives both legal and political scrutiny. The problem is that this path —public regulation, with a democratic mandate and verification mechanisms— is precisely the one the current administration rejects for now. With no public referee and no legal room for agreed self-regulation, the industry is trapped in the worst possible combination in the short term: neither permitted coordination nor a rule to replace it. Whether this lawsuit wins or loses in court, it will likely end up serving, paradoxically, to accelerate pressure on Congress: in the end, whoever decides how to reconcile competition and safety in AI should not be a CEO's essay, but a law.