AI Safety Coordination Could Backfire Legally—Here's Why Companies Are Trapped

Leading AI companies have recently called for a coordinated approach to slowing model development—citing AI agent swarms that have hacked websites, coordinated via secret message boards, and safety warnings from departing engineers. But in seeking this brake on progress, they've created a legal minefield: antitrust regulators may view collective agreements to reduce output as illegal collusion, even if the goal is preventing catastrophic risks.

The Language Problem#

Antitrust law treats corporate communication as evidence. Under the Sherman Act, how executives phrase decisions matters as much as the decisions themselves. Google famously trained staff not to use certain language internally that could suggest anticompetitive behavior, focusing instead on consumer benefits.

When AI leaders use terms like "slowdown" or "pause," they signal something regulators have trained themselves to spot: a possible agreement among competitors to reduce the volume of goods or services entering the market. John Bergmayer of the nonprofit Public Knowledge explains that economists scrutinize whether "you're reducing output"—whether multiple firms have made a pact to "kind of take it easy." The optics look worse than the underlying activity: investing in safety protocols to prevent model misalignment.

Rather than discuss a collective slowdown, Bergmayer suggests, companies could simply emphasize their commitment to joint safety work and present any release delays as a natural consequence of that work. Meta CEO Mark Zuckerberg took this approach, arguing that AI labs have a natural incentive to prevent misalignment because consumers will abandon unsafe products—a competitive, not cartel-like, framing.

What the Law Actually Says#

Several antitrust experts argue the concerns are overblown. David Lawrence, former policy director of the DOJ's Antitrust Division, wrote on LinkedIn that agreements preventing catastrophic risks "increase output and promote competition" and are already protected under the "ancillary restraints doctrine"—a legal principle shielding agreements that are necessary to support legitimate joint conduct.