White House pulls AI review order hours before signing
The White House shelved a voluntary AI review framework on May 21, just hours before the scheduled signing ceremony. Politico had reported the executive order draft on May 20, revealing a system that would have asked developers of advanced AI models to submit their work to federal agencies-including the NSA-up to 90 days before public release.
The order never became official. David Sacks, Trump's AI and crypto czar, contacted the president to relay industry concerns that even a voluntary process could slow American AI companies in competition with China. Trump pulled the order.
What the draft required
The framework was strictly voluntary. Developers of "covered frontier models" would have been encouraged, not mandated, to allow federal review before launch.
The 90-day window would have given agencies time to assess cybersecurity risks and national security implications. The draft explicitly prohibited mandatory licensing or preclearance-companies could still release AI products without government approval.
Why industry pushed back
Tech leaders argued that any review process, even a voluntary one, creates bottlenecks that disadvantage US companies relative to competitors operating under fewer constraints. The argument resonated with the administration's focus on maintaining American AI leadership.
The regulatory context
This executive order would have been the first structured oversight mechanism under the administration's January 2025 order, which aimed to reduce AI restrictions and prevent state-level regulations from fragmenting federal policy.
The tension remains unresolved: national security officials want visibility into powerful AI systems before release. The tech sector wants minimal friction in development pipelines.
What it means for government
If a voluntary review framework cannot survive industry lobbying, substantive federal AI regulation appears unlikely in the near term. Federal agencies currently have limited formal mechanisms to assess frontier AI capabilities before public deployment.
There's a secondary consideration. Countries and organizations making AI purchasing decisions increasingly seek assurances about safety and reliability. A credible review process-even voluntary-could function as a competitive advantage by signaling that US systems meet rigorous standards. Instead, the pullback leaves that market signal absent.
Government officials responsible for AI policy should understand that the current environment offers minimal structured oversight at the federal level. For more on how government agencies are approaching AI governance, see AI for Government.
Your membership also unlocks:
