The Safety Paradox: Why AI Giants Are Treating Antitrust Law as a Shield

### The Coordination Dilemma

In the high-stakes race to develop frontier artificial intelligence, the industry has reached a paradoxical crossroads. As Wired first reported, OpenAI has recently approached members of Congress to seek explicit guidance on a critical question: would it be legal to orchestrate an industry-wide slowdown in AI development?

This inquiry underscores a growing anxiety among AI leaders. While the competitive drive to innovate remains fierce, there is an emerging belief that the pace of development is outstripping the industry's ability to implement necessary safeguards. This tension was highlighted by Jakub Pachocki, OpenAI's chief scientist, who recently published a blog post arguing that coordinating to slow down future development is essential for ensuring the safety of self-improving AI systems. Pachocki suggested that voluntary slowdowns should become commonplace until the industry can establish shared safety bars.

### Antitrust as an Obstacle to Safety

However, the path to a collective pause is blocked by the very laws designed to ensure fair competition. As Wired reports, people close to OpenAI suggest that substantive coordination on safety between AI labs could run afoul of U.S. antitrust statutes, creating a significant barrier to bringing other tech giants into a coordinated effort.

Legal experts have warned that such a move could be interpreted as a restriction of output. In a March article, Nicholas Felstead—currently assistant director of the Australian Competition and Consumer Commission and a former AI policy fellow at the Center for Law & AI Risk—argued that a coordinated pause in development could potentially violate the Sherman Antitrust Act. While Felstead indicated that the legality would depend on the specific details of any agreement, he emphasized that legal uncertainty itself serves as a powerful deterrent, even if the collaborations would eventually survive judicial scrutiny.

### Legislative Responses

Congress has shown early signs of responding to these industry concerns. Wired reports that in July, a bipartisan, bicameral group of lawmakers introduced the "Collaboration on Adversarial Threats and Security Risks Act." This proposed legislation would explicitly allow AI labs to coordinate on security and safety efforts without the threat of antitrust violations.

Caleb Knapp, the director of government affairs at the nonprofit AI Policy Network, endorsed the bill, stating it would establish legal channels for labs to collaborate on security and safety incidents. While Knapp noted a "growing appetite" among lawmakers to address AI safety, he cautioned that the enactment of such laws might be delayed until after the upcoming midterm elections. Currently, the House version of the bill has been referred to the Judiciary Committee but has not yet been taken up.

### The "Fake" Antitrust Defense

Despite the legal concerns raised by OpenAI, some industry figures argue that antitrust fears are a convenient facade. John Schulman, an OpenAI cofounder who now serves as chief scientist at the rival lab Thinking Machines, took to X to argue that citing antitrust as a reason for non-cooperation is "fake." Schulman pointed out that while antitrust laws prohibit certain agreements, they do not prevent companies from jointly developing a proposal. He specifically called on OpenAI and Anthropic to stop feuding and collaborate on a pacing proposal.

Beyond legal liabilities, several other factors complicate the prospect of a coordinated slowdown. Wired reports that AI remains a massive business, with companies competing aggressively for a share of the frontier model market. Furthermore, some executives align with the Trump administration's perspective that maintaining a lead over China is a matter of national security. There is also a fundamental lack of consensus on the best methods for developing safe AI, leading to reluctance among developers to collaborate.

### A Climate of Urgency

The push for a slowdown comes amid a series of alarming events and public warnings. Former OpenAI and Anthropic researcher Jacob Coxon recently issued a stark warning that the current trajectory of AI development puts humanity at risk. This alarm is compounded by actual security failures; Wired reports that OpenAI's agents recently hacked Hugging Face, an incident that illustrates how capabilities are advancing faster than the industry's safeguards.

As these security incidents mount, the industry finds itself trapped in a cycle: the fear of the technology's potential risks creates a desire for collective restraint, yet the rules of the game—specifically antitrust laws—make that restraint a legal liability. For the architects of AI, the question is no longer just about what the technology can do, but whether the law will allow them to stop it in time.