OpenAI Seeks Clarity on the Legality of a Potential AI Industry Slowdown

OpenAI has reached out to members of Congress in recent weeks, seeking explicit guidance on the legality of implementing an industry-wide slowdown in frontier AI advancements, according to sources close to the firm who spoke with WIRED.

Concerns about safety collaboration among AI laboratories might lead to potential violations of antitrust laws, the sources indicate, posing a considerable challenge to achieving buy-in from major tech stakeholders for the initiative.

Recently, OpenAI’s chief scientist, Jakub Pachocki, shared a blog post advocating that the future of AI research should involve “coordinating to slow down future development,” which he believes is crucial for ensuring the safety of self-improving AI systems. In the near term, he anticipates that “voluntary slowdowns will become the norm until common safety standards are implemented.”

However, some legal experts warn that such actions could infringe upon US antitrust legislation. Nicholas Felstead, assistant director of 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 halt in AI advancements could equate to companies limiting output, possibly breaching the Sherman Antitrust Act. The legality would hinge “entirely on the specific details of any agreement,” Felstead noted. “While many safety collaborations might ultimately withstand antitrust evaluations, the uncertainty can be a significant deterrent.”

OpenAI did not provide a comment request before publication.

There are emerging indications that Congress is paying attention. In July, a bipartisan, bicameral assembly of lawmakers introduced a bill titled “Collaboration on Adversarial Threats and Security Risks Act,” explicitly allowing AI labs to collaborate on security and safety efforts without fear of violating antitrust laws. The House version has been sent to the Judiciary Committee but has not yet been addressed.

Caleb Knapp, director of government relations at the nonprofit AI Policy Network, which supports the bill, states it would establish legal pathways for AI labs to unite in addressing safety and security challenges. Knapp expresses that Congress has a “growing appetite to take action” on AI safety, but enacting legislation may need to be postponed until after the forthcoming midterm elections.

While some AI executives may harbor legitimate concerns regarding antitrust issues, another faction of AI leaders contends that such fears serve as a convenient excuse to sidestep real reasons for hesitance in collaboration, which extend beyond potential legal risks.

AI represents a significant business frontier, with companies fiercely competing for a share of the emerging market for advanced models. Some executives echo the Trump administration’s perspective that maintaining an edge over China in AI is vital for national security. Most importantly, varied interpretations among AI developers regarding optimal methods for creating safe artificial intelligence contribute to reluctance in collaboration.

“The first step is for industry leaders OpenAI and Anthropic to cease their conflicts and collaborate on a joint pacing proposal,” John Schulman, an OpenAI cofounder currently serving as chief scientist at the competing lab Thinking Machines, stated in a recent post on X. “They’ll cite antitrust, but that’s misleading—antitrust restricts certain agreements, not the joint development of proposals.”

Long-standing anxieties surrounding the competition among AI companies to develop and release increasingly potent models surged into national focus this summer. Recently, former Anthropic and OpenAI researcher Jacob Coxon heightened the prevailing concerns, issuing a stark public alert about the risks posed by AI developers to humanity. A series of security incidents in recent months, including OpenAI’s agents breaching Hugging Face, revealed the inadequacies of the industry’s safeguards in keeping pace with advancing model capabilities. This sequence of events has prompted numerous lawmakers to urgently advocate for AI regulation.

This is an edition of Maxwell Zeff’s Model Behavior newsletter. Read previous newsletters here.