Three camps in Trump administration clash over AI regulation after executive order collapse
President Donald Trump scrapped an AI executive order on May 21, exposing internal divisions over how to regulate the technology. The decision came hours before he was scheduled to sign the order, following a last-minute call from former AI czar David Sacks.
The White House contains three distinct factions with competing visions for AI policy, according to two senior administration officials.
The deregulation camp
Sacks leads the group pushing for minimal regulation. He argues that excessive oversight could hamper the U.S. AI industry's ability to compete with China. Sacks contacted Trump directly to block the order, citing industry concerns that it would be too burdensome for a relatively new sector.
The security hawks
Defense Secretary Pete Hegseth and undersecretary Emil Michael, a former Silicon Valley executive, want stricter controls on advanced AI models. They fear that powerful systems could be exploited by foreign adversaries or used for cyberattacks on critical infrastructure.
Their concerns intensified after Anthropic released Mythos in April-a highly advanced model capable of finding and exploiting unknown security flaws in IT systems.
The middle ground
Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent propose a voluntary oversight framework. Under their approach, AI companies would allow the U.S. government to review new models before public release, without mandatory compliance.
A senior White House official described the framework: "It wasn't government telling these companies what they could and couldn't do, but it requested that the U.S. government get a first look at any new models, just to be sure that they couldn't be exploited by bad actors."
How the order was built
The administration began drafting AI policy in February, initially focused on cybersecurity standards and NSA monitoring of federal networks. Work accelerated after Mythos' public arrival in April.
In late April, the Office of the National Cyber Director held confidential meetings with OpenAI, Google, and Anthropic. Industry representatives signed non-disclosure agreements before reviewing the proposed language.
Two table reads occurred on May 20 at the Eisenhower Executive Building. AI executives reviewed a hard copy of the order and appeared to accept its terms. Trump was prepared to sign the following morning.
Then Sacks intervened. "David just thought better of it and contacted the president and said, 'I think you're making a mistake,'" the official said.
What happens next
The order is postponed, not dead. Administration officials are exploring revised versions that could still reach Trump's desk.
One option under discussion would give the government 90 days to review new models before release. Another would remove the voluntary pre-release requirement entirely, allowing companies to make their own commitments outside the executive order.
The remaining sections would establish a Treasury-led AI cybersecurity clearing house to coordinate software vulnerability patching across the industry.
White House spokesperson Liz Huston said the administration is "committed to addressing the challenges posed by this emerging technology" while maintaining innovation leadership.
For government officials working on AI policy, these internal divisions signal that the regulatory framework remains unsettled. Understanding the competing priorities-competitiveness, national security, and industry flexibility-will be essential as policy develops. Learn more about AI governance and policy frameworks or explore AI for Government resources.
Your membership also unlocks:
