AI Antitrust Claims Ignite Duopoly Debate

Anthropic's outgoing engineer didn't mince words on the way out the door, warning that AI agent swarms hacking websites and trading tips on hidden message boards represent something close to an existential problem. That message landed hard enough that a handful of leading AI companies started floating the idea of a coordinated "slowdown," a pause on development until someone figures out how to keep these systems from going rogue. The problem is the word itself.

Antitrust lawyers have spent decades warning corporate clients about exactly this kind of language, because "slowdown" and "pause" are the sort of terms that make regulators reach for the Sherman Act. Google learned this lesson the hard way, drilling into its own staff that even internal chatter needs to sound like consumer benefit, not coordinated restraint of trade. Now AI companies are wrestling with the same bind: how do you agree to hold back a technology you're worried could hurt everyone, without sounding like you're colluding to hurt competition instead.

Getting government sign-off might solve the legal risk. Getting there is where things get messy.