OpenAI asks Congress if an industry-wide frontier-AI slowdown would violate antitrust law
A bipartisan safe-harbor bill would permit cross-lab safety coordination, but it is stalled in House Judiciary.
OpenAI has been privately asking members of Congress whether coordinating an industry-wide slowdown in frontier AI development could be legal under US antitrust law. The outreach lands as lawmakers weigh a safe-harbor bill that would explicitly protect certain AI safety and security collaboration from antitrust risk.
Key Takeaways
- OpenAI has asked members of Congress in recent weeks whether orchestrating an industry-wide slowdown in frontier AI development would be legal under US antitrust law.
- Substantive cross-lab safety coordination is being treated internally as an antitrust risk, complicating any attempt to bring major tech companies into a shared pacing effort.
- OpenAI chief scientist Jakub Pachocki publicly argued for “coordinating to slow down future development,” saying he expects “voluntary slowdowns to become commonplace until shared safety bars are established.”
- A July bipartisan, bicameral safe-harbor bill would explicitly permit AI labs to coordinate on security and safety work without antitrust exposure, but the House version has not moved after referral to the Judiciary Committee.
OpenAI’s Quiet Antitrust Check-in: Can Labs Coordinate a Frontier-AI Slowdown?
OpenAI has asked members of Congress in recent weeks for guidance on whether an industry-coordinated slowdown in frontier AI development would be legal, according to people close to the company. The question is narrow, but the implication is broad: if the legal perimeter is unclear, any serious cross-lab “pacing” plan risks dying in counsel review before it ever becomes a public proposal.
People close to OpenAI described substantive coordination on safety between AI labs as a potential antitrust problem, and as a practical obstacle to bringing major tech companies into any slowdown effort. That matters for markets because “frontier AI” is not a slogan here. It is the most advanced models near the state of the art, and the cadence of their releases has become a macro input into tech sentiment that often bleeds into crypto risk appetite.
OpenAI did not respond to a request for comment ahead of publication, leaving the scope of its outreach and the specific guidance it sought undisclosed.
Why a “Voluntary Slowdown” Runs Into Sherman Act Output-Restriction Risk
The legal tension is mechanical. US antitrust law prohibits certain forms of coordination that reduce competition, and coordinated output restrictions are a classic red flag. A “voluntary slowdown” can look like safety work in a blog post and like a production cap in a complaint, depending on what the labs actually agree to do.
Nicholas Felstead, an assistant director at the Australian Competition and Consumer Commission and a former AI policy fellow at the Center for Law & AI Risk, argued in a March article that a coordinated pause in AI development may amount to companies restricting output and could violate the Sherman Antitrust Act. His key qualifier was that legality would depend “entirely on the precise details of any agreement,” which is the part that turns this from an abstract debate into a gating item for executives.
Felstead also flagged a second-order constraint that traders tend to underestimate: uncertainty itself. “But even if most safety collaborations would ultimately survive antitrust scrutiny, legal uncertainty can act as a powerful deterrent.” Even a defensible coordination structure can be too expensive to attempt if the downside is years of litigation risk or a regulator deciding the agreement is really a market-allocation scheme with safety branding.
That is why OpenAI’s outreach reads less like a philosophical exercise and more like pre-clearance. If the goal is to get multiple labs, and potentially major tech firms, to move in the same direction on pacing, the agreement design becomes the product.
The Safe-Harbor Attempt: What the Collaboration on Adversarial Threats and Security Risks Act Would Change
Congress has a live, if stalled, legislative path to reduce that agreement-design risk. In July, a bipartisan, bicameral group of lawmakers introduced the “Collaboration on Adversarial Threats and Security Risks Act,” which would explicitly permit AI labs to coordinate on security and safety work without risking antitrust violations.
The House version was referred to the Judiciary Committee and has not been taken up. That procedural status is the whole point for near-term expectations: without movement, the market is left with ambiguity rather than permission, and ambiguity tends to push coordination into informal channels that are harder to verify and easier to dispute.
Caleb Knapp, director of government affairs at the nonprofit AI Policy Network, which endorsed the bill, said it would create legal channels for AI labs to work together to address safety and security incidents. Knapp also said Congress has a “growing appetite to get something done” on AI safety, while cautioning that passing anything into law may have to wait until after the upcoming midterm elections.
The politics and the law are now entangled with the industry’s own messaging. OpenAI chief scientist Jakub Pachocki wrote last weekend that the best path forward includes “coordinating to slow down future development,” and that he expects “voluntary slowdowns to become commonplace until shared safety bars are established.” At the same time, John Schulman, an OpenAI cofounder now chief scientist at rival AI lab Thinking Machines, argued on X earlier this week that antitrust is being used as cover: “First step is for industry leaders OpenAI and Anthropic to stop feuding and work on a pacing proposal together,” he wrote, adding, “They'll cite antitrust, but that's fake—antitrust prohibits certain agreements, but not from jointly developing a proposal.”
Signals Traders Can Monitor: Hill Movement, Lab Messaging, and Any Formal Pacing Proposal
The cleanest signal is procedural, not rhetorical. Any House Judiciary Committee action on the “Collaboration on Adversarial Threats and Security Risks Act” would reduce the odds that pacing coordination stays stuck in a gray zone, whether that action is a hearing, markup, or re-referral.
The second signal is whether frontier labs move from general calls for coordination to a concrete pacing proposal with defined scope, duration, and enforcement. Pachocki’s framing normalizes “voluntary slowdowns,” but antitrust exposure is driven by the details Felstead emphasized, not by the intent.
Third, informal guidance can matter even absent legislation. Public statements from lawmakers or regulators that sketch how antitrust would treat AI safety coordination could change corporate risk tolerance, even if they do not bind future enforcement.
Finally, the dispute itself is a tell. If more industry leaders publicly echo or rebut Schulman’s critique that antitrust is a pretext, it raises the odds that the next phase is a public, structured proposal rather than quiet bilateral conversations.
My Read: Policy Uncertainty Is Becoming Part of the Frontier-AI “Release Cycle”
The threshold that matters is whether pacing becomes a documentable, counsel-approved coordination structure rather than a moral argument. OpenAI’s outreach to Congress suggests the company is treating antitrust uncertainty as a gating item for any cross-lab slowdown, and Felstead’s framing explains why: a coordinated pause is easy to reframe as an output restriction if the agreement is even slightly sloppy.
If the House safe-harbor bill stays stuck, the setup looks more like a sentiment catalyst than a fundamental shift, because the industry will keep talking about “voluntary slowdowns” without a legally durable way to do them together. This only starts to matter in practical terms when either Congress creates a safe harbor or a major lab publishes a pacing proposal whose details survive the Sherman Act output-restriction test.
