OpenAI wants to know if AI industry slowdown will even be legal

OpenAI asked Members of Congress have been seeking clear guidance in recent weeks on whether it would be legal to orchestrate an industry-wide slowdown on advanced AI developments, people close to the company told WIRED.

Significant security coordination among artificial intelligence labs could lead to antitrust violations, the people said, posing a major obstacle to getting big tech giants to join the effort.

Over the weekend, OpenAI chief scientist Jakub Paczocki published a blog post arguing that the best path forward for the world of AI research involves “coordinating the slowdown of future development,” which he believes will be key to ensuring the safety of self-improving AI systems. In the short term, he expects “voluntary slowdowns to become commonplace until common safety barriers are in place.”

But some legal scholars say such efforts could run afoul of U.S. antitrust laws. Nicholas Felstead, assistant director of the Australian Competition and Consumer Commission and former AI policy fellow at the AI ​​Law and Risk Center, argued in a March article that a coordinated pause in AI development could lead to companies limiting product output, potentially breaching the Sherman Antitrust Act. That would “depend entirely on the precise details of any agreement,” Felsted wrote. “But even if most security cooperation projects ultimately survive antitrust scrutiny, legal uncertainty could be a powerful deterrent.”

OpenAI did not respond to a request for comment before publication.

There are some early signs that Congress is listening to the issue. In July, a bipartisan, bicameral group of lawmakers introduced a bill called the Cooperation Against Adversarial Threats and Security Risks Act, which would explicitly allow artificial intelligence labs to coordinate security efforts without risking violating antitrust laws. The House version has been referred to the Judiciary Committee but has not yet been considered.

Caleb Knapp, director of government affairs for the nonprofit AI Policy Network, which supported the bill, says it would create legal channels for artificial intelligence labs to work together to resolve security incidents. Knapp adds that there is “a growing desire in Congress to do something” on AI safety, but says any legislation may have to wait until the upcoming midterm elections.

While some AI leaders may have genuine concerns about antitrust, another camp of AI leaders argue that these concerns are just a convenient cover to avoid discussing the real reasons why AI developers might be wary of collaboration, which go well beyond potential legal obligations.

First, artificial intelligence is a huge business, and these companies are competing fiercely for share of the nascent market for cutting-edge models. Some leaders share the Trump administration’s view that outpacing China in artificial intelligence is critical to national security. Perhaps most importantly, different AI developers have very different opinions about how best to develop safe artificial intelligence, and so may be reluctant to work with each other on this issue.

“The first step is for industry leaders OpenAI and Anthropic to stop their feud and work together on the pace proposal,” wrote John Shulman, an OpenAI co-founder who is now chief scientist at rival AI lab Thinking Machines, in a post on X earlier this week. “They will cite antitrust laws, but this is fake: antitrust laws prohibit certain agreements, but not joint development of proposals.”

This summer, long-standing concerns about a race among artificial intelligence companies to create and release ever more powerful models came into the national spotlight. This week, former Anthropic and OpenAI researcher Jacob Coxon added to the growing alarm with a stark public warning that AI developers are putting humanity at risk. In recent months, a number of security incidents, including the hack of Hugging Face by OpenAI agents, have demonstrated that industry security measures have not kept pace with the growing capabilities of models. The sequence of events has prompted many lawmakers to urgently call for regulation of AI.

This edition Maxwell Zeff Behavior Model Newsletter. Read previous newsletters Here.