WASHINGTON — In a quiet yet high-stakes maneuver that underscores the growing existential anxiety within the artificial intelligence sector, OpenAI has recently approached members of the United States Congress to explore a controversial legal question: Could a pact among rival AI labs to slow down development violate federal antitrust laws?
According to sources close to the discussions who spoke with WIRED, the ChatGPT maker’s recent outreach on Capitol Hill is part of a broader, increasingly urgent internal and external campaign. The initiative follows public calls from OpenAI Chief Scientist Jakub Pachocki for the industry to establish shared safety standards and voluntary pauses. However, the path toward industry-wide collaboration is fraught with legal landmines, fierce commercial competition, and the compounding nature of AI advancement itself.
The inquiry into antitrust exemptions highlights the paradox defining the modern generative AI boom. While top researchers and executives increasingly warn that moving too fast could unleash unmanageable societal—or even existential—risks, the multi-billion-dollar race for market dominance penalizes any individual company that dares to slow down on its own.
Main Facts: The Dilemma of Speed Versus Safety
The core tension in the current AI landscape centers on a classic collective action problem. On one side, industry leaders and independent watchdogs warn that competitive pressures are driving firms to bypass critical guardrails. On the other side, the economic and geopolitical incentives to build the next frontier model are simply too lucrative to ignore.
- The Antitrust Barrier: Under U.S. law, direct coordination among competitors to restrict output, delay product releases, or fix prices can easily trigger devastating civil and criminal antitrust investigations by the Department of Justice (DOJ) or the Federal Trade Commission (FTC). OpenAI’s outreach to lawmakers is an effort to understand if Congress would be willing to carve out safe harbors or enact legislation protecting safety-driven collaborations.
- The Safety Backlash: Recent events have brought these safety concerns into sharp relief. Most notably, Jacob Coxon, a former engineer at rival lab Anthropic, publicly resigned this week, citing profound fears regarding the existential risks posed by unbridled AI development. Coxon claimed on social media that the people building these systems "earnestly believe that it could kill us all by the end of the decade," adding that executives often downplay these fears publicly while expressing genuine terror behind closed doors.
- The Legislative Landscape: Lawmakers are beginning to take notice. Bipartisan legislation introduced by Senators Adam Schiff and Jim Banks aims to protect certain national security-related collaborations among tech firms, provided they give advance notice to the DOJ. Yet, a broader exemption allowing commercial rivals to legally collude on slowing down model deployment remains uncharted territory.
Chronology: How the AI Safety Debate Escalated
The friction between commercial acceleration and safety caution has evolved rapidly over the past year, marked by shifting corporate policies, legislative delays, and whistle-blower exits.
- May 2025: The geopolitical stakes were thrown into sharp focus when U.S. President Donald Trump delayed a planned executive order on artificial intelligence over concerns that strict federal mandates might weaken America’s technological lead over China.
- June 2025: President Trump ultimately signed a revised executive order. The directive established a voluntary review process for advanced models prior to their public release, attempting to balance national security concerns with the breakneck pace of domestic innovation.
- February 2026: Cracks in industry-wide self-regulation began to show publicly when both OpenAI and Anthropic quietly dialed back some of their foundational safety language. Anthropic Chief Science Officer Jared Kaplan defended the move, arguing that slowing down development unilaterally made little strategic sense while competitors continued to push forward.
- August 2026: Demonstrating that internal alarms were still firing, OpenAI temporarily paused internal work on its "Astra" AI model due to critical cybersecurity concerns and a lack of stronger protective safeguards. However, the pause was treated as an isolated internal adjustment rather than an industry-wide standard.
- September 2026: The debate boiled over into the open. OpenAI Chief Scientist Jakub Pachocki called for a voluntary industry slowdown until developers can definitively prove their systems are safe. Shortly thereafter, OpenAI initiated its quiet legal probe of Congress regarding antitrust laws, culminating in the high-profile resignation of Anthropic engineer Jacob Coxon.
Supporting Data: The Mechanics of the AI "Arms Race"
To understand why voluntary slowdowns have largely failed, industry experts point to the unique economic and technical architecture of modern artificial intelligence.
Miranda Bogen, chief technologist at the Center for Democracy and Technology, explained to Decrypt that the structural incentives in the AI sector actively punish caution.
"Commercial and geopolitical competition in the AI space is incredibly intense, leading to a concerning dynamic where companies are incentivized to release products before their risks are fully understood," Bogen said. "Even when internal staff knows more research and testing is needed, their companies are facing immense pressure to cut corners and skip critical safety tests, despite evidence piling up about the consequences of moving too fast."
Duncan Sabien, head of communications at the Machine Intelligence Research Institute (MIRI), emphasized that financial rewards compound the problem exponentially. Every major technological breakthrough unlocks billions in new venture capital and enterprise funding, cementing the market power of the winner.
Furthermore, Sabien pointed out a compounding technical reality: artificial intelligence is uniquely self-reinforcing.
"This is especially true since gains in intelligence are compounding; each new system makes it easier to train and deploy the next, more-advanced system," Sabien noted. "Sans some sort of coordination mechanism, stepping back just means the other guy gets a lead."
This dynamic explains why individual exits or isolated corporate pauses rarely achieve lasting change. Sabien argued that individual whistle-blowers accomplish little if they are simply replaced by slightly less cautious engineers. True reform, he suggested, requires a critical mass of researchers and executives who share a common understanding and manage to break through the coordination barrier together.
Official Responses and Stakeholder Perspectives
As lawmakers weigh OpenAI’s informal queries, opinions across the policy, legal, and technical communities are sharply divided regarding how—or if—the federal government should intervene.
- OpenAI: By approaching Capitol Hill, OpenAI is signaling that it recognizes the limits of voluntary, unilateral restraint. Without legislative cover or antitrust relief, any formal pact among OpenAI, Anthropic, Google, and other major labs to halt or throttle the training of frontier models could invite immediate prosecution for monopolistic collusion.
- Civil Society and Ethics Advocates: Organizations like the Center for Democracy and Technology argue that market forces alone cannot solve the crisis. They contend that external regulatory pressure is desperately needed to protect the public from deployment milestones driven by venture capital rather than empirical safety audits.
- The Talent Pool: The resignation of Jacob Coxon highlights a growing internal crisis of conscience within elite AI labs. Coxon’s public assertions that top-tier researchers routinely express private dread over their own creations challenge the corporate narrative that current safeguards are fully adequate.
Implications: Where Does the Industry Go From Here?
The intersection of antitrust law and artificial intelligence safety represents a novel frontier for American jurisprudence. If Congress chooses to explore exemptions that allow AI labs to coordinate on safety milestones or deployment pauses, it would represent a historic shift in how antitrust law interacts with nascent, high-risk technologies.
However, crafting such legislation is fraught with political peril. Lawmakers must navigate the fine line between enabling responsible safety protocols and inadvertently granting major tech incumbents a legal shield to freeze out smaller competitors, stifle open-source innovation, and cement an oligopoly.
Meanwhile, the geopolitical shadow of the U.S.–China tech rivalry continues to loom large. Any perception that American AI labs are slowing down their development cycles—even for safety reasons—runs directly into national security arguments that cede ground to foreign competitors.
For now, OpenAI’s probe of Congress serves as a stark acknowledgment that the current trajectory of artificial intelligence is running on borrowed time. Unless a viable mechanism for cooperative restraint can be forged through legal, legislative, or cultural breakthroughs, the relentless march toward more powerful AI systems will likely continue unchecked, driven by a competitive dilemma that no single company feels it can afford to lose.
