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

OpenAI has asked Congress a simple but loaded question: if the biggest AI companies agree to slow down together, is that legal? The company wants to know whether coordinated efforts to pause or throttle development of the most powerful AI models would break antitrust laws — the rules that stop companies from secretly teaming up to control markets. It's an odd situation: a company asking Washington for permission to go slower.

Here's the legal knot. Under the Sherman Antitrust Act, when competing companies agree to limit how much they make or release, that can be treated as an illegal output restriction. One lab can decide to pause on its own — that's perfectly fine. But rivals making a mutual pledge to hold back? That invites federal scrutiny, even if their stated reason is safety. This legal uncertainty may itself discourage companies from coordinating, experts say.

The timing is striking because AI leaders rarely agree on anything. Anthropic CEO Dario Amodei called for slowing the pace of AI improvement. OpenAI's Sam Altman quickly agreed publicly. Elon Musk said Amodei is right. But skeptics — including Rep. Josh Gottheimer and OpenAI co-founder John Schulman — say the antitrust worry is a convenient excuse. If these CEOs are truly worried, they can slow down today, on their own. Schulman called the antitrust argument 'fake,' noting the law doesn't stop companies from jointly proposing rules.

A bipartisan bill would create a legal safe harbor for safety collaborations, but it's stuck in committee ahead of midterm elections. Meanwhile, critics warn that a government-approved slowdown framework could lock in today's dominant players as a cartel — keeping smaller rivals out while the biggest labs control the pace. For regular people, this shapes how fast new AI tools arrive, what they can do, and which companies get to build them.