Leading AI companies have called for a coordinated "slowdown" in AI development following reports of AI agent swarms hacking websites and coordinating via secret message boards, but antitrust experts say the firms' language itself may trigger regulatory scrutiny, according to a new analysis published by Wired. The companies have expressed worry that pausing development would violate antitrust laws, yet legal scholars note that while the charged rhetoric isn't helping their case, preventing rogue killer AI development likely aligns with the Sherman Act's goals of promoting competitive markets.
Under antitrust law, how employees at a company discuss business choices often carries as much weight as the decisions themselves. Google famously instructed its staff to avoid certain phrases—even in internal communications—that might suggest anticompetitive conduct, instead directing them to highlight how decisions would enhance products and benefit customers. From a regulatory perspective, terminology like "a slowdown" or "a pause" may raise more red flags than the underlying activity: building safeguards to keep advanced AI models from going rogue. John Bergmayer, legal counsel for nonprofit Public Knowledge, notes that regulators typically examine whether companies are cutting back on output—essentially making a deal to "kind of take it easy." A collectively agreed-upon slowdown lacking any clear purpose could be read by officials as an anticompetitive pact to limit trade. Meta CEO Mark Zuckerberg, whose firm recently avoided a major FTC antitrust lawsuit, weighed in by refusing to endorse an explicit "slowdown" at all, arguing instead that AI labs have a "strong natural incentive" to improve AI agent behavior because consumers reject models that do things people don't intend, and companies failing to achieve proper alignment "will fall behind" in the market.
The analysis explains that agreements preventing catastrophic risks actually "increase output and promote competition" and are already shielded under law by the "ancillary restraints doctrine," according to David Lawrence, a former policy director at the DOJ's Antitrust Division. A career FTC antitrust attorney commented that "no humanity would result in no competition." Roger Alford, a professor at Notre Dame Law School and former second-in-command for the DOJ Antitrust Division, points out that collectively choosing not to adopt safety measures could expose AI labs to accusations of "quality fixing"—when companies mutually decide not to enhance their own products. He references a European antitrust case where automakers collaborated on emissions-reducing technology but agreed not to compete on improvements beyond legal requirements, ultimately paying roughly a billion-dollar fine. Industries can already reduce antitrust liability through the National Cooperative Research and Production Act of 1993, which allows them to establish a standards-development organization as long as they notify the FTC and DOJ, Bergmayer notes.
Trustbusters typically view antitrust exemptions skeptically, saying they end up enlarging big players and blocking newer firms from gaining market traction. Yet many employees at major AI companies have genuinely raised alarms about how rapidly their workplaces are releasing new models, potentially sacrificing safety. Bergmayer observes that sometimes companies seek regulation to pull up the ladder behind them, but other times "it really is that they feel like the market is pressuring them to act in a way that they think they shouldn't." Both Anthropic and OpenAI filed confidential IPO paperwork this summer and hold valuations near or beyond a trillion dollars each. Anthropic is expected to go public next month, while OpenAI CEO Sam Altman said his company would postpone its IPO until 2027 due to recent safety worries. The Ramp AI Index, which tracks AI adoption rates, indicates the two firms' models have comparable usage rates and are competing neck and neck for the top position. Alford suggests the companies want an exemption because "they don't want to unilaterally disarm while the others keep going at a breakneck speed." Political pressure compounds the challenge: after news of the slowdown proposal, President Donald Trump posted that the government already has "tremendous CRIMINAL and REGULATORY power over these companies" and declared "WHOEVER WINS AI, WINS!"
If the federal government launched an antitrust investigation into an AI slowdown, it would be classified as a "conduct investigation." Unlike merger investigations, which impose strict time limits on government probes, a conduct investigation can stretch for years—AI companies could be required to produce millions of pages of documents, executives and key employees could be pulled into depositions, and hundreds of other workers could have their devices subject to litigation holds, even if the government ultimately decides it lacks a case. With no new regulations on the horizon and likely no antitrust exemption forthcoming, frontier AI labs will have to write their own rules and potentially dodge investigations along the way. The choice between safety protocols and competitive velocity may ultimately rest with individual companies willing to accept market risk, rather than coordinated industry action that invites legal jeopardy.