President Donald Trump pulled back an artificial intelligence policy order two days ago after citing concerns that the draft executive order would have "inhibited" the AI industry and could have hurt U.S. competitiveness with China.
"I was hearing concerns, but I was also seeing the concerns myself," Trump said Friday morning. "I have concerns about it, and I don't want to approve anything until it's done properly. I want the industry to be able to continue to win, we're leading by a lot over China and everybody else, and I want to continue, and I felt it was inhibiting the industry."
The Shelved Policy Framework
The draft policy would have created a voluntary oversight system in which developers of advanced AI models could submit their products for review by federal agencies as much as 90 days before release. The policy was being discussed for advanced AI models like Mythos.
Industry representatives and White House officials expressed concern that such reviews could slow the rapidly evolving industry. More significantly, stakeholders worried that the voluntary vetting system could one day become mandatory—a regulatory ratchet that would lock in compliance burdens even as technology and market conditions changed.
David Sacks, the president's former AI czar, emerged as one of the most vocal opponents of the executive order. According to senior White House officials and people familiar with the matter, Sacks argued the reviews could slow the industry and hamper the United States' ability to compete with China. He conveyed his concerns to the president in the hours leading up to Thursday's scheduled signing ceremony. "It's truly David on a mission," one White House official said. Industry leaders also opposed the order, according to people familiar with the discussions.
Administration Divided on Approach
Notably, National Cyber Director Sean Cairncross and Treasury Secretary Scott Bessent—both lead administration officials on the issue—were not informed of the postponement until after Trump had already decided to pull the order. A White House official said, "Sean isn't the problem."
Senior administration officials including Bessent and National Economic Council Director Kevin Hassett have repeatedly emphasized the need for companies and government agencies to shore up cyber defenses against advanced AI models with the capacity to find holes in security systems faster than humans ever have. This creates a tension within the administration: the desire to address genuine cybersecurity vulnerabilities while avoiding regulatory structures that could disadvantage American companies in global competition.
Industry Response and Path Forward
White House officials and industry representatives said they still expected some policy to emerge from the Trump administration, but the executive order would now head back to the drawing board and be reworked. One tech industry lobbyist said, "I don't think it's dead. I think that there will be an effort to make some changes and get some sort of a framework in place, if for nothing else to address the cyber issues."
Another industry representative described the process as "a mess" and said it "wouldn't surprise them" if the executive order ended up pulled altogether.
Some top tech leaders said Friday they had not tried to stop the order's signing. Elon Musk posted on X, "I still don't know what was in that EO and the President only spoke to me after declining to sign." Meta CEO Mark Zuckerberg's company spokesperson Andy Stone said, "Mark didn't speak to the president until after the event had already been canceled."
Industry representatives said tech companies were mostly coalescing around the need to address cyber risks posed by new models. Anthropic and OpenAI have released models with advanced capabilities but not yet released them to the public, and other AI companies are expected to produce them too. One industry representative with direct knowledge of policy negotiations said tech companies were "pretty much OK" with the executive order, though they had lingering questions about which agencies would oversee the processes. That representative said, "It's chaos, but for us, we still feel that we need to do something on cyber."
Why This Matters:
The decision to pause the AI policy order reflects a fundamental tension in regulatory strategy: how to address legitimate security concerns without imposing compliance burdens that disadvantage American innovation in a sector critical to national competitiveness. Trump's pullback suggests the administration recognizes that regulatory frameworks, once established, tend toward expansion rather than restraint. The fact that Sacks and other industry voices successfully argued against the order demonstrates the power of evidence-based criticism of regulatory overreach. However, the administration's commitment to addressing cybersecurity risks through some form of policy framework—even if reworked—indicates this issue will not disappear. The challenge ahead is designing a system that protects genuine security interests without creating the regulatory infrastructure that could eventually become mandatory, undermining the competitive advantages American AI companies currently hold over China and other rivals.