President Donald Trump called off a planned signing ceremony for a new executive order on artificial intelligence Thursday, May 21, telling reporters at the White House that the order’s text gave him concerns about America’s competitive edge in the AI race.
The Oval Office event, scheduled for Thursday afternoon, was expected to bring together executives from leading US AI companies — including Anthropic, OpenAI, and Google — to mark the launch of what would have been one of the most consequential tech policy actions of Trump’s second term. Instead, the signing was postponed indefinitely.
“We’re leading China, we’re leading everybody, and I don’t want to do anything that’s going to get in the way of that lead,” Trump told reporters, according to coverage from the Associated Press distributed through outlets including WWL.
What the Order Would Have Done
The order would have established a framework for the federal government to vet the national security risks of the most advanced AI systems before their public release. It was designed as a voluntary collaboration with US-based companies including Anthropic, OpenAI, and Google.
That structure marked a notable shift from the administration’s earlier posture. Trump’s January 2025 Executive Order 14179, titled “Removing Barriers to American Leadership in AI,” was specifically designed to roll back the AI safety regulations established by his predecessor and clear runway for rapid industry growth. The $500 billion Stargate infrastructure project, announced shortly after, doubled down on that approach by directing massive federal-backed investment toward AI computing capacity.
The proposed Thursday order would have moved in a different direction — adding a pre-release government review layer for the most capable models. Trump’s decision to pull it back, after seeing the final text, suggests that internal debate within the administration has not yet resolved how aggressively the federal government should screen AI systems.
Why the Sudden Pause
The pressure for some form of government review has been building inside the administration for months. Treasury Secretary Scott Bessent and outgoing Federal Reserve Chair Jerome Powell convened an urgent meeting with Wall Street CEOs in April to warn them about cybersecurity risks posed by Anthropic’s AI model Claude Mythos. Bessent later spoke about the meeting at CNBC’s “Invest in America Forum” in Washington, describing the model as “very powerful” and emphasizing that banks needed to align on best practices for handling AI-related cyber risk.
Those warnings, combined with growing concern about AI’s ability to identify cybersecurity vulnerabilities in software systems, helped drive the push for some kind of structured federal review. But the administration has also been wary of any action that could be interpreted as throttling the very companies it has spent the past year championing.
Serena Booth, a computer science professor at Brown University and former AI policy fellow on a Democratic-led Senate committee, told the AP that the pattern reflects “competing factions” inside the administration. “We do see this kind of public fighting,” she said. “‘We will release an executive order. No, we won’t. We’re going to sign it this afternoon. Oh, the signing is canceled.’ I think this whiplash is because we’re seeing these fractures.”
Implications for State AI Laws
The other dimension of the order, reported earlier by Axios in December 2025, involved federal pushback against state-level AI legislation. State laws have proliferated quickly: Utah’s disclosure rules have been in force since May 2025, with $2,500 penalties per violation; Texas implemented sweeping plain-language disclosure requirements on January 1, 2026; Colorado’s AI Act enforcement begins June 30, 2026; and California has advanced its own transparency rules covering generative AI developers.
For state legislatures and the businesses that operate across them, Thursday’s delay extends a period of regulatory uncertainty. Companies deploying AI in healthcare, hiring, financial services, and consumer products are currently navigating a state-by-state patchwork with no clear federal framework. The longer that patchwork remains the dominant reality, the more companies must build compliance infrastructure designed to satisfy the strictest state — typically California or Colorado — rather than a unified national standard.
What the Administration Is Saying Publicly
At a White House press briefing Tuesday, Vice President JD Vance declined to discuss specifics from the draft order but framed the administration’s broader stance as a balance. “The president wants us to be pro-innovation. He wants us to win the AI race against all other countries in the world,” Vance said, adding, “We also want to make sure that we’re protecting people.”
Some screening of advanced AI systems is already happening. The Commerce Department announced earlier in May that it had signed agreements with Google, Microsoft, and Elon Musk’s xAI to evaluate their most powerful models before public release — agreements that build on previous arrangements made under the Biden administration with Anthropic and OpenAI. That announcement was later removed from the Commerce Department’s website, an unexplained reversal that fits the broader pattern of administrative whiplash on the issue.
Dean Ball, a former White House tech policy adviser and lead author of Trump’s AI policy roadmap who is now at the Foundation for American Innovation, told the AP that the disagreements reflect what he characterized as “healthy tension” inside an administration genuinely conflicted about how to engage frontier AI companies on cybersecurity. “Ultimately, I’m fine with them taking time to get this right,” Ball said.
When that resolution comes, and what it looks like, remains an open question.
