President Donald Trump's last-minute decision to shelve a planned artificial intelligence executive order has exposed a growing divide inside the White House over how far the federal government should go in regulating AI, according to a new report.
POLITICO recalled that Trump abruptly scrapped the order on May 21, just hours before it was expected to be signed, after former AI czar David Sacks intervened directly with the president. The move derailed months of internal work and reopened a high-stakes debate over whether Washington should keep a light touch on AI or impose stronger guardrails as models being released to the public become increasingly stronger.
The dispute has split senior officials into three broad camps, the outlet said. Sacks and allies favor a less restrictive approach, arguing that too much regulation could weaken American companies in the race against China.
On the other side stand Defense Secretary Pete Hegseth and Undersecretary Emil Michael, who have pushed for tougher barriers around advanced models because of fears they could be exploited by rivals or bad actors.
A third group, described as closer to the middle, includes White House chief of staff Susie Wiles and Treasury Secretary Scott Bessent. They have supported a voluntary framework under which AI companies would give the federal government an early look at new models before they are released publicly.
The order that nearly reached Trump's desk would not have imposed a mandatory approval process. Instead, it would have created a voluntary system for companies to consult with the government on their most advanced models, allowing officials to assess whether the technology could be used to attack critical infrastructure, discover cyber vulnerabilities, or aid foreign adversaries.
A senior White House official told the outlet that the draft was not intended to let the government dictate what companies could release. Rather, the official said, it requested that the administration get "a first look" at new models to make sure they could not be exploited by "bad actors."
White House spokesperson Liz Huston pushed back on the idea that the administration is divided, telling POLITICO that Trump's team is "united" behind a policy that balances innovation with security. "Under President Trump's leadership, America is winning the global AI race and leading the world in innovation," Huston said.
The internal clash intensified after Anthropic's release of Claude Mythos Preview, an advanced cybersecurity-focused AI model that has alarmed officials because of its ability to find and exploit flaws in computer systems.
The U.K. AI Security Institute said in April that its evaluation found Mythos Preview represented a "step up" over previous frontier models in cyber capabilities. That kind of power has sharpened the central question facing the White House: How does the United States stay ahead of China without letting cutting-edge AI become a tool for cyberattacks?
Industry pressure also played a role. The Washington Post reported that Silicon Valley figures warned the proposed order could slow innovation and become a de facto regulatory regime, even if framed as voluntary. Trump later said he postponed the order because he did not like parts of it.
For now, the order is not dead. POLITICO reported that officials are considering several options, including revised language that could allow the government up to 90 days to review new models before public release, or removing that provision altogether and focusing instead on a Treasury-led AI cybersecurity clearinghouse.
