top of page

The Trump AI Executive Order Got Pulled Hours Before Signing: Why Musk, Zuckerberg, and Sacks Won That Phone Call

  • May 27
  • 2 min read

On May 21, 2026, the White House canceled the signing ceremony for a sweeping AI oversight executive order hours before the Oval Office event was scheduled to begin. The cancellation is the most consequential AI policy story of the month and one of the clearest recent examples of direct CEO lobbying influencing an interagency AI policy process. This post lays out what the order would have done, who pushed to stop it, and what the cancellation tells us about where federal AI policy actually goes from here.


What the Order Would Have Done

The order would have established a voluntary mechanism for AI developers to engage with federal agencies and submit advanced models for security review up to 90 days before public release. It included no licensing regime, no mandatory hold period, and no obligation for any private lab to participate. National security agencies were reportedly expected to participate in portions of the review process. In practice, this was the lightest-touch version of pre-launch federal oversight that has been proposed in any serious draft over the past 18 months.


Who Was On the Phone

Per Semafor and Axios reporting, three people spoke directly with President Trump between Wednesday night and Thursday morning: Elon Musk of xAI, Mark Zuckerberg of Meta, and venture capitalist David Sacks, who had recently served as Trump's AI and crypto czar. The Washington Post's account adds that the lobbying was last-minute and that Sacks personally objected to the regulatory framing. Sources familiar with the discussions described strong opposition inside the administration to the regulatory framing of the proposal.


OpenAI reportedly supported the order. That is a meaningful split in the industry coalition. The companies on the phone with the President opposed a voluntary, NSA-overseen review window; the company behind one of the most widely used frontier AI platforms backed it.

The Public Framing

The official reason for the cancellation centers on competition with China. The argument is that any federal review window, even voluntary, signals to capital and talent that frontier development in the United States carries regulatory drag and that the marginal lab will choose a jurisdiction without that drag. The counterargument inside the executive branch was that voluntary engagement plus NSA evaluation is a cheap insurance policy against catastrophic deployment surprises. The cancellation aligned more closely with the first argument.


What This Tells Us About the Next 12 Months

Federal AI policy will likely live in NIST standards work rather than executive orders. The NIST agent standards initiative was already the most active federal channel; the May 21 cancellation makes it the dominant one. Builders who want to track where the binding rules actually appear should watch NIST drafts, not Oval Office signing schedules.

State-level activity will accelerate. The federal vacuum is the strongest single driver of the 600-plus state bills already filed in 2026. Companies operating in California, New York, and Texas should plan for a heterogeneous compliance regime through at least 2027.

Industry lobbying will be more visible. The May 21 episode reinforced how much direct CEO access can shape AI policy outcomes. The next round of policy proposals will be drafted with that knowledge in the room.


Sources


 
 
bottom of page