There is a lawyerly problem hiding inside the group hug. When Anthropic, OpenAI and others all publicly agree to slow the pace of AI development at the same time, they may be showing admirable restraint, or they may be doing something competition regulators have a very specific word for. As Bloomberg flagged this week, the industry-wide “let’s all ease off” moment raises real antitrust questions, and the companies calling for it are exactly the market leaders a pause would protect.
Why the timing is legally awkward
Antitrust law is relaxed about a company choosing to go slower on its own. It gets twitchy when competitors coordinate, explicitly or with a nod and a wink, on how fast they will develop or release products. “We should all agree to hold back the frontier” sounds noble in an essay. Translated into competition law, an agreement among dominant rivals to restrain output or innovation is close to the textbook definition of the thing regulators exist to stop.
And look at who benefits. A frontier freeze locks in the current order, with today’s leaders on top and challengers denied the chance to leapfrog by pushing harder. Whether or not anyone intends it that way, “everyone please stop racing” from the front-runners has the same practical effect as pulling up the ladder. That is precisely the pattern that draws regulatory scrutiny, regardless of how sincere the safety worries are.
The takeaway
None of this means the safety concerns are fake. It means the labs are caught between two contradictory pressures: the safety case wants coordinated restraint, and antitrust law forbids coordinated restraint among dominant competitors. Squaring that circle almost certainly requires governments to set the rules, so no single company (and no informal club of them) is deciding the pace. Until then, every joint call for a “slowdown” is going to attract two audiences: people who find it reassuring, and lawyers who find it interesting. (Not legal or investment advice.)
Related: the slowdown call itself.