OpenAI Asks Congress: Can AI Labs Team Up to Slow Down?

OpenAI has spent recent weeks asking members of Congress a surprisingly simple question: if AI companies agreed to slow down together, would that be legal? The company's chief scientist, Jakub Pachocki, wrote in a blog post that "coordinating to slow down future development" is the best path for AI safety, and that voluntary slowdowns should become normal until everyone agrees on shared safety standards. In plain terms, OpenAI is floating the idea of a handshake deal to pump the brakes — and wants to know if Washington would object.

The problem is antitrust law. These are the rules that stop competing companies from colluding — say, all the airlines agreeing to fly fewer planes so tickets stay expensive. A coordinated AI slowdown could look a lot like that: companies deliberately limiting their own output. Legal scholars say it would depend entirely on the fine print, but that even the uncertainty can scare companies off. If your lawyers can't promise you're safe, you don't sign.

Congress is starting to listen. In July, a bipartisan group of lawmakers introduced a bill that would explicitly let AI labs coordinate on security and safety without antitrust risk. It's sitting in committee, and insiders say nothing will pass until after the midterms. Not everyone buys the legal worry. Some AI leaders call antitrust a convenient excuse, pointing out that these companies are fierce rivals, disagree about safety, and many want to beat China first. One former OpenAI cofounder put it bluntly: joint proposals are legal, so stop using lawyers as a shield.

The stakes are getting real. This summer brought public warnings from AI researchers and reports of AI agents — software that acts on its own — hacking into systems, showing that safeguards lag behind capability. Whether or not Congress acts, the question of who gets to slow AI down, and whether anyone can, will shape the tools showing up in your job, your inbox, and your kids' classrooms.