The White House came within hours of signing a new executive order that would have created a voluntary review process for advanced AI models. Then it didn’t.

Politico reported on a draft of the AI executive order on May 20, revealing a framework that would have required developers of “covered frontier models” to submit their technologies for federal agency review, including by the National Security Agency, up to 90 days before public release. The signing was scheduled for May 21. It never happened.

What the draft actually said

The draft outlined a voluntary system, not a mandatory one. Developers of the most advanced AI systems would have been asked, not forced, to let federal agencies take a look under the hood before shipping products to market.

The 90-day pre-release review window would have given agencies like the NSA time to assess cybersecurity implications and national security risks.

Crucially, the draft explicitly prohibited mandatory licensing or preclearance requirements. Companies wouldn’t need government permission to launch AI products, but they’d be encouraged to let Washington peek at the most powerful ones first.

Why it got pulled

The executive order was shelved just hours before the signing ceremony. David Sacks, Trump’s former AI and crypto czar, contacted the president on May 21 to relay concerns from industry leaders. Their argument was straightforward: even a voluntary review process could create bottlenecks that slow down American AI companies at precisely the wrong moment in the global competition with China.

Trump reportedly grew uncomfortable with provisions he felt could hinder US competitiveness. So he pulled the plug.

The broader regulatory backdrop

The administration issued an executive order on January 23, 2025, aimed at reducing restrictions on US AI leadership and pushing back against state-level regulations that could fragment the national approach to AI governance.

The now-postponed executive order represented the first real attempt to introduce any structured oversight mechanism under this framework. On one side, there are legitimate national security concerns about releasing increasingly powerful AI systems without any government visibility into their capabilities. On the other, there’s a tech sector arguing that any friction in the development pipeline hands advantages to competitors operating under fewer constraints.

What this means for investors

A voluntary review framework is about as light-touch as government oversight gets. If even that can’t survive contact with industry lobbyists, it tells you something about the likelihood of more substantive regulation arriving anytime soon. For now, AI companies are operating in what amounts to a regulatory free zone at the federal level.

There’s also the competitive angle to consider. The argument for pulling the executive order was that it would slow US companies relative to Chinese rivals. But the counterargument is that countries and enterprises making purchasing decisions about AI systems increasingly want assurances about safety and reliability. A credible review framework, even a voluntary one, could have functioned as a competitive advantage rather than a handicap.