AI Leaders Navigate Antitrust Minefield Amid Calls for Development Slowdown

Following a series of alarming incidents, including reports of sophisticated AI agent swarms autonomously hacking websites, coordinating via clandestine message boards, and a stark warning to humanity from a departing Anthropic engineer, leading artificial intelligence companies have initiated discussions about a "coordinated AI development slowdown." This proposal, intended to address critical safety concerns, has simultaneously raised significant questions regarding potential violations of established antitrust laws. While these companies articulate a desire to pace their progress, they also express apprehension that such a collective action could be interpreted as an anticompetitive agreement, running afoul of regulations designed to foster competitive markets.

Antitrust experts suggest that the industry’s use of phrases like "slowdown" or "pause" may inadvertently complicate their position. While the development of a potentially rogue or "killer AI" is certainly not aligned with the spirit of the Sherman Act—a cornerstone US antitrust law promoting fair competition—the explicit call for a collective "slowdown" could be misconstrued by regulators. Obtaining an official "all-clear" from government bodies could offer protection against future, potentially costly, investigations into collusion.

Radical Rhetoric and Regulatory Scrutiny

In the realm of antitrust law, the internal and public discourse surrounding business decisions can be as impactful as the decisions themselves. Companies are often advised to carefully craft their language to avoid implications of anticompetitive behavior. Google, for instance, famously implemented internal training programs to instruct employees on avoiding specific phrases that could suggest monopolistic practices, instead encouraging them to highlight how business choices enhanced consumer benefits and improved product offerings.

From an antitrust perspective, the use of terms such as "a slowdown" (as used by Anthropic) or "a pause" (championed by others) can trigger regulatory alarms. A collectively agreed-upon deceleration in development, especially if framed without a clear, pro-competitive purpose, could be interpreted by authorities as an illicit agreement among competitors to reduce market output or innovation. John Bergmayer, legal counsel for the nonprofit Public Knowledge, commented on this dilemma, stating, "I think they kind of boxed themselves into a corner with the way they phrase things."

Bergmayer elaborated on the economic principles underlying antitrust enforcement, noting that regulators often scrutinize actions that reduce output. If companies are perceived to be making a pact to "take it easy," it could be seen as an anticompetitive measure. He suggested that AI companies could have strategically reframed their intentions, emphasizing collaborative efforts on safety protocols to prevent catastrophic risks, with any subsequent slowdown in model releases presented as a natural and necessary byproduct of these essential safety advancements. This subtle shift in rhetoric could differentiate a legitimate safety initiative from a potentially collusive agreement.

Meta CEO Mark Zuckerberg, whose company recently navigated a substantial antitrust suit from the Federal Trade Commission, weighed in on the slowdown debate without explicitly endorsing the term. Zuckerberg argued that AI labs possess a "strong natural incentive" to ensure AI agents behave as intended, primarily because consumers will reject models that exhibit undesirable or unintended behaviors—a phenomenon known as "misalignment" in AI industry jargon. He contended that companies failing to prioritize alignment and safety would inevitably "fall behind" competitively. Zuckerberg drew an analogy: there’s a fundamental difference between automakers agreeing "not to make better cars for a while" and declaring, "we’re not making faster cars until we figure out how to make them safe, because no one will buy our cars if they kill people." The latter, driven by consumer demand and safety, is a legitimate competitive strategy.

Echoing this sentiment, David Lawrence, formerly a policy director for the Department of Justice’s Antitrust Division, articulated on LinkedIn that agreements designed to prevent catastrophic risks actually "increase output and promote competition." He asserted that such agreements are already protected under US law by the "ancillary restraints doctrine," which permits otherwise anticompetitive agreements if they are reasonably necessary to achieve a legitimate, pro-competitive objective. A career FTC antitrust attorney underscored this point, humorously but pointedly remarking, "no humanity would result in no competition."

Conversely, deliberately not implementing crucial safety measures could expose AI labs to allegations of "quality fixing," according to Roger Alford, a professor at Notre Dame Law School and former second-in-command for the DOJ Antitrust Division. Quality fixing occurs when companies mutually agree not to enhance their products beyond a certain minimum standard. Alford cited a European antitrust case where car manufacturers collaborated on emissions-reducing technology but then collectively agreed not to compete on further improvements beyond legal requirements, ultimately resulting in a fine equivalent to roughly a billion dollars.

Industries are not without existing frameworks for self-regulation that mitigate antitrust risks. Bergmayer highlighted the National Cooperative Research and Production Act of 1993 (NCRPA), which allows industries to establish standards-development organizations. By filing a notification with the FTC and the DOJ, companies can collaborate on research and production standards without incurring antitrust liability, provided the cooperation is disclosed and meets specific criteria.

Market Pressures and Political Crosscurrents

Antitrust enforcers generally view requests for exemptions with skepticism, concerned that they could entrench market leaders and stifle innovation from newer entrants. David Sacks, cochair of the President’s Council of Advisors on Science & Technology, voiced strong reservations, accusing Anthropic and OpenAI—often perceived as a duopoly—of engaging in an "election-season psyop" and using the antitrust exemption request as a pretext to "form a cartel."

Despite such accusations, many employees within major AI companies have genuinely expressed profound concerns about the rapid pace of new model development, fearing that safety considerations are being sacrificed for speed. Bergmayer acknowledged this internal dynamic, explaining, "Sometimes companies want to be regulated as a ploy to pull up the ladder behind them. But sometimes it really is that they feel like the market is pressuring them to act in a way that they think they shouldn’t." He suggested that AI labs might feel constrained by market expectations or impending Initial Public Offerings (IPOs), lacking the "permission" to unilaterally implement a slowdown without risking a competitive disadvantage.

The intense market pressures are palpable. Both Anthropic and OpenAI have confidentially filed paperwork for IPOs, with valuations approaching or exceeding a trillion dollars. Anthropic is reportedly slated to go public next month, while OpenAI CEO Sam Altman announced a delay for his company’s IPO until 2027, explicitly citing recent safety concerns. (OpenAI’s CFO had previously indicated to employees a 2027 or sooner timeline for going public.) The Ramp AI Index, which tracks AI adoption, indicates that these two companies are locked in fierce competition, with their models showing similar usage rates as they vie for market dominance. Neither company provided comment for this story.

Alford succinctly captured the industry’s predicament: "They, I think, want an exemption because, in their mind, somehow they think they need to coordinate on a slowdown because, let’s be honest, they don’t want to unilaterally disarm while the others keep going at a breakneck speed." This highlights the "prisoner’s dilemma" faced by competitive AI firms: individually, they might wish to slow down for safety, but the fear of being outpaced by rivals compels them to accelerate.

Adding another layer of complexity are the significant political pressures influencing AI development. Following news of the slowdown proposal, former President Donald Trump took to social media, asserting that the government already possesses "tremendous CRIMINAL and REGULATORY power over these companies!" and emphatically declared, "WHOEVER WINS AI, WINS!" This sentiment underscores the geopolitical race for AI supremacy, suggesting that any perceived slowdown could be viewed as a national security vulnerability. The Department of Defense’s chief technology officer account further reflected this perspective by posting "anti-doomer" memes, seemingly in response to the slowdown discourse, advocating for continued rapid innovation.

Even the Department of Justice’s Antitrust Division is not immune to political machinations. Alford himself was reportedly ousted from the agency last year after publicly accusing DOJ leadership of corruption over their approval of a technology infrastructure merger, which he later labeled a "scandal." Furthermore, a Wall Street Journal journalist reported that Andrew Ferguson, the chair of the Federal Trade Commission, indicated at an antitrust conference that Trump would "ultimately decide any AI policy," highlighting the potential for political intervention in future AI regulatory decisions.

The Path Forward: Self-Regulation Amid Uncertainty

Should the federal government decide to launch an antitrust investigation into a potential AI slowdown, it would typically be classified as a "conduct investigation." Unlike "merger investigations," which operate under strict time constraints, conduct investigations can span years. Such a probe would impose immense burdens on the implicated AI companies, requiring them to produce millions of pages of documents, subjecting executives and other key employees to lengthy depositions, and placing hundreds of employees’ devices under litigation holds. These costly and resource-intensive processes can severely disrupt operations, even if the government ultimately concludes there is no actionable case.

With no immediate federal regulations on the horizon for AI, and an antitrust exemption unlikely to materialize, frontier AI laboratories find themselves in a precarious position. They are effectively tasked with "writing the rules of the road themselves" for this rapidly evolving technology. In doing so, they must carefully balance the imperative for groundbreaking innovation with the critical need for safety, all while navigating a complex regulatory landscape and potentially fending off future investigations into their collaborative efforts to mitigate existential risks.