DOJ Signals Openness to AI Safety Coordination as Antitrust Guidance Review Begins

On September 17, 2026, Associate US Attorney General Stanley Woodward told an audience at Fordham University's 53rd Annual Conference on International Antitrust Law and Policy in New York that coordination among artificial intelligence companies on cybersecurity and safety does not strike him as anticompetitive. His remarks, delivered during a period of intense debate over how AI labs should handle safety risks, signal that the Justice Department may be open to allowing competitors to work together on security issues without fear of antitrust enforcement.

The comments come after Anthropic CEO Dario Amodei called on the US government for a narrow antitrust waiver that would let competing AI labs coordinate on slowing development and setting common safety standards. In a 3,800-word essay published in mid-September 2026, Amodei wrote that the industry must slow the pace at which it improves AI models, and he committed Anthropic to embedding third-party evaluators who would have access similar to internal risk staff.

The debate has been sharpened by a regulatory vacuum. In December 2024, the Justice Department withdrew the 2000 Antitrust Guidelines for Collaboration Among Competitors and has not replaced them. That document had provided a roadmap for companies trying to determine where cooperation ends and collusion begins. Without it, firms have faced uncertainty about whether joint safety efforts could trigger antitrust scrutiny.

Not everyone in Washington is welcoming the idea of relaxed antitrust scrutiny for AI safety collaboration. FTC Chairman Andrew Ferguson expressed skepticism on September 15 toward AI firms that simultaneously request new regulations and antitrust exemptions, warning that such standards could end up locking out smaller competitors. On the same day, a Senate Judiciary hearing featuring FBI Director Kash Patel saw Senators Ted Cruz and Josh Hawley reject antitrust exemptions. The DOJ's position thus sits in contrast to the FTC's more cautious stance.

Key Facts

Woodward said that existing guidelines spell out the extent to which companies can coordinate on hacking risks without running afoul of rules prohibiting anticompetitive behavior. He confirmed that the Justice Department is reviewing whether to update its existing guidance on cybersecurity cooperation to cover AI. Bloomberg Law reported on September 17 that Woodward said the Trump administration is considering whether to update interagency antitrust guidance on cybersecurity to address new threats posed by artificial intelligence. The Washington Examiner reported on September 17 that Woodward said coordinating on cybersecurity or safety is not obviously anticompetitive, and that companies seeking to self-regulate do not need DOJ involvement.

Despite the public debate, Woodward noted that no frontier AI lab had actually contacted his office to discuss antitrust concerns. 'It doesn't occur to me that coordinating on cybersecurity or security is anticompetitive,' he said. He added that the Antitrust Division is open to meeting with companies, but that firms at the forefront of development had not requested a meeting. When asked whether companies need a waiver to coordinate, Woodward said that to the extent companies are looking to regulate themselves or fix problems, they should do that, but working with the DOJ is unnecessary.

CryptoBriefing reported on September 17 that the DOJ is signaling it is open to letting AI companies work together on safety without triggering antitrust alarms. The report noted that Amodei's proposal is notable for its specificity: he is not asking for broad antitrust immunity, only permission to coordinate on the pace and safety of AI development. Amodei warned that an agent swarm could take over the internet with a persistent botnet within six to twelve months, costing hundreds of billions of dollars. He pointed to a swarm that had breached Hugging Face as an example of the risks.

Noah News reported on September 18 that Woodward's remarks come as the department builds out a wider AI policy framework. The department has published internal and public materials on how it uses AI in areas ranging from criminal justice to national security, and its latest AI inventory shows a substantial increase in recorded use cases. At the same time, the DOJ has shown willingness to take an aggressive line when it believes AI regulation crosses constitutional or competition boundaries. In April, the DOJ intervened in xAI's challenge to Colorado's Algorithmic Discrimination law, arguing the measure unlawfully compelled companies to build in discriminatory constraints. The department also sued Apple in March 2024 over alleged smartphone monopolisation, and in June 2025 required Safran to divest assets in an aerospace deal.

Opposition from some quarters remains firm. The Next Web reported on September 13 that Amodei's essay committed Anthropic to embedded third-party evaluators and asked Washington for a narrow antitrust waiver, but much of what he proposes is already law in the European Union. Since 2 August 2025, the EU AI Act's systemic risk rules for general-purpose models have required companies including Anthropic to assess loss-of-control risks. European officials, including tech chief Henna Virkkunen, have said the same is not true globally. More than a thousand people at AI labs had asked for a pacing mechanism in July, according to The Next Web.

Analysis

Woodward's comments suggest that the Justice Department is drawing a line between coordination on security and safety, which it views as potentially procompetitive or at least not anticompetitive, and traditional collusion on prices or output. What this really means is that the DOJ is not granting a blanket exemption for AI companies to coordinate on all matters, but it is signaling that safety-focused collaboration is unlikely to face immediate enforcement action. The department appears to be taking a wait-and-see approach, preferring voluntary self-regulation over formal waivers.

The bigger picture here is that the withdrawal of the 2000 guidelines has left a void that the DOJ is now considering filling with updated guidance tailored to AI. This would be a significant shift, as it could provide companies with clarity while maintaining antitrust oversight. However, the FTC's skepticism, voiced by Chairman Ferguson, highlights an interagency split. The FTC worries that allowing dominant AI labs to set safety standards could entrench their market power and disadvantage smaller competitors who cannot afford to comply with costly standards.

Amodei's call for an antitrust waiver is also notable because much of what he proposes is already mandated in the European Union. The EU AI Act's systemic risk rules have been in effect since 2 August 2025, requiring assessments of loss-of-control risks. The absence of similar requirements in the US has created a regulatory gap that Amodei and others argue could be filled by voluntary commitments, but only if antitrust law does not stand in the way. The DOJ's openness may be an attempt to keep pace with global regulatory developments without imposing heavy-handed rules.

Yet the political dynamics are complicated. Senators Cruz and Hawley rejected antitrust exemptions at a September 15 hearing, reflecting a populist concern that big tech companies could use safety rhetoric to stifle competition. The DOJ's aggressive antitrust enforcement against Apple and its intervention in the xAI case show that the department is not averse to taking on major tech firms. Balancing these priorities will be delicate. Woodward's statement that companies do not need DOJ involvement to self-regulate suggests the department prefers a hands-off approach, but it may face pressure to act if coordination leads to tangible harms.

Why It Matters

The outcome of the DOJ's review could shape how AI companies approach safety. If the department formally updates its guidance to cover AI, it could encourage labs to share information about vulnerabilities and best practices, potentially reducing risks from malicious actors. But it could also lead to coordination that slows innovation or raises barriers to entry. The FTC's warning about locking out smaller competitors is a reminder that antitrust law exists to protect competition, not just to punish bad actors.

Globally, the US is lagging behind the EU in AI regulation. The EU AI Act's systemic risk rules have been in place since August 2025, and European officials have pointed out that companies like Anthropic are already subject to them. If the US does not provide clear guidance, AI labs may face a patchwork of rules or may relocate to jurisdictions with more favorable treatment. Amodei's proposal even mentions an eventual agreement with Beijing, underscoring the international dimension of AI safety coordination.

Until the DOJ finalizes its guidance, companies will operate in a gray area. Woodward's remarks provide some comfort, but they are not a formal policy. The FTC's skepticism and Congressional opposition mean that any move to relax antitrust scrutiny will be contested. The debate reflects a broader tension between the need for rapid AI safety measures and the risk of entrenching the power of a few dominant firms.

Next Up

The Justice Department is expected to continue its review of whether to update its cybersecurity cooperation guidance to cover AI. Companies may seek meetings with the Antitrust Division, though so far none have. The FTC's stance and Congressional hearings will also influence the outcome. Sens. Cruz and Hawley have already signaled their opposition to antitrust exemptions, and more hearings are likely.

Lawmakers may consider legislation to clarify the boundaries of antitrust law for AI safety collaboration. Meanwhile, the EU's experience with the AI Act will serve as a reference point. How the US balances safety and competition will set a precedent for other emerging technologies.