OpenAI Queries US Congress on Sherman Act Implications for Coordinated Industry AI Slowdown

WASHINGTON — OpenAI has approached members of the U.S. Congress to clarify whether orchestrating a voluntary, industry-wide development slowdown among leading artificial intelligence laboratories would violate federal antitrust laws, according to reporting by WIRED. The inquiry directly tackles an unexpected obstacle in AI governance: the fear that competing developers working together to halt or pace the training of frontier models could be treated under the Sherman Antitrust Act as an illegal cartel restricting output and suppressing innovation.

Background

The frontier AI sector is locked in an aggressive capability race. Industry leaders such as OpenAI, Google DeepMind, and Anthropic are investing tens of billions of dollars to scale neural networks, chasing human-level performance and autonomous agentic workflows. However, as frontier models demonstrate increasingly sophisticated reasoning, planning, and self-improving capabilities, concerns over catastrophic safety failures, systemic misuse, and loss of control have moved from theoretical discussions into operational realities.

Historically, the core dynamic preventing any single lab from pausing high-risk research has been a classic prisoner's dilemma: if one firm pauses model training to reinforce safety evaluations, rivals could leapfrog it in capability, capturing commercial dominance and market share. While individual companies can unilaterally slow their own development, doing so alone carries steep competitive penalties.

To solve this collective action problem, safety advocates have long proposed multilateral pacing commitments or joint safety thresholds. However, entering into such pacts creates legal friction. The Sherman Antitrust Act of 1890, the bedrock of U.S. competition law, strictly prohibits agreements among horizontal competitors that limit production, restrict output, or stifle commercial activity. In ordinary commercial markets, rival firms agreeing to restrict development would face scrutiny from the Federal Trade Commission (FTC) or Department of Justice (DOJ) as unlawful collusion.

What Happened

OpenAI has quietly held conversations with congressional lawmakers to determine whether the legal framework governing anti-competitive agreements would penalize labs that agree to coordinate safety pauses, according to reporting detailed by NeoTeo and WIRED.

The policy outreach arrives amid shifting internal stances at OpenAI. The lab's chief scientist, Jakub Pachocki, published a blog post outlining that frontier AI developers should consider "coordinating to slow down future development" until industry-wide safety and alignment benchmarks are established, as reported by Northeast Times. Pachocki emphasized that no laboratory has yet fully resolved fundamental alignment and monitoring challenges required to scale continuously at maximum velocity.

Simultaneously, OpenAI Chief Executive Sam Altman told employees internally that the firm could consider slowing its pace of frontier model deployment alongside peer institutions, while acknowledging that securing unanimous industry participation would be difficult, according to Bloomberg and Gate News. Safety concerns have been underscored across the ecosystem by recent internal agent breaches, prominent researcher resignations, and a July petition signed by over 1,000 AI researchers demanding global mechanisms to pace automated research.

Legal experts note that OpenAI’s concerns are grounded in statutory reality. Nicholas Felstead, assistant director of the Australian Competition and Consumer Commission and former fellow at the Center for Law & AI Risk, noted in an analysis for Lawfare cited by ForkLog that joint developer pauses could legally mirror an output restriction pact. Even if a court ultimately determined that a safety pause served public welfare under the antitrust "rule of reason," the threat of protracted litigation and treble damages acts as a chilling deterrent.

Why It Matters

OpenAI’s congressional query underscores a critical collision between twentieth-century competition statutes and twenty-first-century frontier technologies.

For the AI ecosystem, the query reveals the structural limits of self-regulation. Without formal safe harbors or statutory clarity, lab executives face legal exposure if they attempt to craft self-imposed pacing frameworks. Conversely, critics and industry observers worry that granting broad antitrust immunity could give dominant incumbents an anticompetitive shield. If market leaders are legally permitted to coordinate pacing around proprietary "safety bars," smaller open-source developers or emerging competitors could find themselves boxed out by standards defined strictly by incumbent labs.

Furthermore, not all technical leaders agree that antitrust law is a genuine barrier. John Schulman, an OpenAI co-founder and chief scientist at rival lab Thinking Machines, argued that antitrust rules prohibit specific market-rigging schemes while still permitting companies to collaboratively formulate technical standards and safety specifications, as noted by Northeast Times. The debate highlights divergent views on whether labs need new statutory protections or whether existing legal frameworks are simply being used to test political appetite for market stabilization.

What's Next

Attention is now shifting toward Capitol Hill, where lawmakers are weighing whether to grant explicit antitrust exemptions for AI safety governance.

In July, a bipartisan group introduced the Collaboration on Adversarial Threats and Security Risks Act, sponsored in the Senate by Jim Banks and Adam Schiff and in the House by Bob Latta and George Whitesides, according to AI Weekly. The bill was initially drafted as a narrow antitrust safe harbor allowing AI companies to share threat intelligence and defend against foreign national security threats, such as model distillation from Chinese competitors.

Policy analysts will be closely monitoring whether the House Judiciary Committee amends or expands this legislation to include broader safe-harbor provisions covering synchronized capability freezes and shared pacing bars. Meanwhile, regulators at the DOJ and FTC have yet to issue formal policy statements clarifying their enforcement posture on cooperative safety agreements.

Unless Congress legislates explicit legal safe harbors, any attempt by frontier AI labs to coordinate development timelines will remain caught between escalating calls for safety governance and the strictures of U.S. antitrust law.