The Ceremony That Wasn't: How Industry Insiders Derailed Trump's AI Executive Order
Everything was set for a high-profile Oval Office moment. Silicon Valley's elite had been summoned. A landmark executive order on artificial intelligence governance — months in the making — sat ready for signature. Then, hours before the ceremony, it fell apart.
On May 21, President Donald Trump abruptly canceled the signing of what would have been his administration's most significant move yet on AI oversight, citing a geopolitical argument his advisers had deployed to devastating effect: anything resembling regulation, he warned, could cost America its lead over China.
"We're leading China, we're leading everybody, and I didn't want to do anything to get in the way of that lead," Trump told reporters in the Oval Office. "I really thought it could have been a blocker."
The last-minute reversal exposed deep fractures inside the administration — between those who see AI safety measures as prudent national security infrastructure and those who view any oversight as an invitation for competitors to close the gap.
What the Order Would Have Done
The shelved order was not a sweeping regulatory framework. It would have directed federal agencies to work with leading AI companies on a voluntary basis to test advanced "frontier" models up to 90 days before public release. The order had a significant cybersecurity component, tasking the Cybersecurity and Infrastructure Security Agency, the Office of the National Cyber Director, and the Office of Science and Technology Policy with developing standards to identify which models warranted review.
A second section would have formalized efforts to strengthen national security and Defense Department systems against AI-assisted cyberattacks and would have supported critical infrastructure protection — including utility companies and rural hospitals.
The urgency was not abstract. Anthropic's Mythos Preview model had recently demonstrated the ability to autonomously discover thousands of severe cybersecurity vulnerabilities in major operating systems — capabilities so alarming that Anthropic declined to release it publicly. Vice President JD Vance, speaking the day before the canceled ceremony, had framed the effort in sweeping terms: "Right now, we're working in a collaborative way with the technology companies, and we're just trying to make sure that the American people are as safe as possible," he said, warning that "a bad actor could use Mythos to target various cybersecurity vulnerabilities."
Who Pushed Back — and Why
Behind the scenes, the opposition was swift and well-connected. According to Axios, Trump's AI adviser David Sacks — the venture capitalist and former White House AI czar — along with Meta CEO Mark Zuckerberg and xAI CEO Elon Musk, spoke with the president between Wednesday night and Thursday morning. The companies did not respond to requests for comment.
The case against signing was ideological as much as procedural. "The main reason why the executive order signing was delayed was because 'he just hates regulation,'" a source familiar with the discussions told Axios. The order was characterized as "unnecessary" — "just something doomers wanted," the source added.
A separate tech industry source raised a structural objection: the draft had given the Treasury Department an unusually prominent coordinating role in cybersecurity vulnerability response — a function historically handled by CISA and NIST. "It's not clear just objectively speaking why Treasury is involved and what is their substantive expertise in this area," the source told Axios. There were also concerns that a 90-day window — far longer than the two weeks industry preferred — could prevent companies from sharing models with allied nations conducting their own reviews.
The Broader Federal-State Battle
The canceled order arrived within a broader policy fight over who governs AI in America. A December 11, 2025 executive order — "Ensuring a National Policy Framework for Artificial Intelligence" — had established what the administration called a "One Federal Standard" doctrine, arguing that because AI is inherently interstate commerce, states should be preempted from imposing their own safety, bias-mitigation, or reporting requirements. That order targeted statutes like the Colorado AI Act, created a DOJ AI Litigation Task Force, and conditioned federal broadband funding on states' willingness to pause conflicting AI enforcement.
The cancellation of the May order revealed that even within the administration, consensus on how far to push AI governance had not been achieved. Those who had been pushing for safety measures in Washington — including some inside the White House's own National Cyber Director's office — had welcomed the EO as a long-overdue signal. With it gone, that work stalled.
A Narrower Order, Quietly Signed
About two weeks after the failed ceremony, on June 2, Trump signed a substantially revised version of the executive order — privately, without the planned photo op. The new order, "Promoting Advanced Artificial Intelligence Innovation and Security," scaled the voluntary pre-release review window from 90 days to 30 days and included an explicit carve-out: "Nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, preclearance, or permitting requirement for the development, publication, release, or distribution of new AI models, including frontier models."
Scientific American described the final, signed order as representing "a major shift in the Trump administration's approach, which had been more laissez-faire" — notable, the outlet observed, given that Trump had so recently backed away from an earlier version over the same safety concerns.
What to Watch Next
The underlying tensions are far from resolved. The Office of the National Cyber Director has signaled it is working on additional AI security initiatives beyond the signed EO. State legislatures continue advancing AI governance bills despite the December 2025 preemption framework, and legal challenges to that order are widely expected.
For those who hoped the original, more expansive draft would become law, the renegotiated version is a half-measure. For accelerationists, it is precisely the right outcome: light-touch, voluntary, and careful not to hand Beijing any rhetorical ammunition. The question is whether the next model that demonstrates dangerous capabilities — another Mythos, perhaps — will force the debate open again.
"Right now, we're working in a collaborative way with the technology companies, and we're just trying to make sure that the American people are as safe as possible."— JD Vance, Vice President of the United States