AI Companies' 'Slowdown' Plea Runs Into Antitrust Trouble
Leading AI firms, including Anthropic and OpenAI, have called for a coordinated slowdown in AI development over safety fears, but such an agreement could clash with US antitrust law. Legal experts say the wording, not the intent, is the real risk.
Following reports of AI agent swarms hacking websites and coordinating secretly, along with a stark warning from a departing Anthropic engineer, several major AI companies have called for a coordinated development "slowdown." At the same time, those same companies have voiced concern that such coordination could violate US antitrust laws.
Antitrust experts say the core issue lies less in the idea itself and more in how it's phrased. Terms like "slowdown" or "pause" can sound, in an antitrust context, like an agreement to restrict market output — a hallmark of anticompetitive behavior. John Bergmayer of the nonprofit Public Knowledge says the companies could instead have framed this as a shared effort to develop safety protocols, treating any slowdown as an incidental side effect.
Zuckerberg's Alternative Framing
Meta CEO Mark Zuckerberg, whose company recently avoided a major FTC antitrust suit, stopped short of endorsing an explicit "slowdown." He argued AI labs already have a strong natural incentive to fix misbehaving models, since consumers won't want agents that act unpredictably, and companies that fail to get this right will lose ground to competitors.
Others warn the opposite risk: if AI labs jointly agreed not to improve safety at all, they could face accusations of "quality fixing" — companies colluding to withhold improvements from consumers. A past European case saw automakers pay roughly a billion-dollar fine for a similar arrangement limiting emissions-technology improvements.
Market and Political Pressure
Both Anthropic and OpenAI filed confidential paperwork this summer for initial public offerings, with valuations near or above a trillion dollars each. Anthropic is expected to go public next month, while OpenAI's CEO said its IPO would be delayed until 2027 over safety concerns.
President Donald Trump posted that the government already holds "tremendous criminal and regulatory power" over these companies. If federal regulators launched a formal antitrust probe into a coordinated slowdown, it could stretch on for years, demanding millions of pages of documents and depositions from employees. With no new regulation or antitrust exemption in sight, AI labs may be left to set their own rules.
