On Sept. 12, Anthropic chief executive Dario Amodei called for industry-wide cooperation to slow AI advances in favour of stronger safety measures. OpenAI’s Sam Altman, SpaceXAI’s Elon Musk and Google DeepMind co-founder Demis Hassabis publicly backed the proposal, according to a lawsuit that casts their alignment as far more than a policy debate.¹

Amodei himself acknowledged the legal sensitivity, suggesting the US government could mediate or enable cross-lab safety discussions and issue a narrow waiver for them. Altman, meanwhile, said OpenAI supported a federal framework with consistent safety requirements, but did not think companies needed to wait for an antitrust exemption or legislation before beginning safety work.¹

Less than a week later, paid users of ChatGPT, Claude, Grok and Gemini sued in the US District Court for the Northern District of California. Their proposed nationwide class action says subscribers paid for access to increasingly capable models, only to face an alleged agreement that diminished what they received. “An agreement that slows improvement lowers the quality of what subscribers receive for the price they pay,” the complaint says.²

The plaintiffs do not argue that companies must race ahead recklessly. Their point is narrower—and potentially more consequential: every firm may set its own safety thresholds, but rival firms cannot jointly decide how quickly competing products improve. The complaint calls such coordination a restriction of output, conduct barred by Section 1 of the Sherman Act.²

The political backdrop sharpens the divide. President Donald Trump has resisted calls to limit AI, arguing regulation could damage the sector, while promising an AI task force and an “AI czar.” Sen. Josh Hawley has likewise rejected any antitrust exemption for the major labs.¹ David Sacks amplified the same warning in reposting a Wall Street Journal opinion line: an AI antitrust exemption “would invite collusion.”David Sacks (DavidSacks)³

For now, the case turns on whether public agreement over catastrophic-risk safeguards was merely parallel advocacy—or the beginning of a private pact that subscribers say made their products worse.