President Donald Trump abruptly canceled a White House ceremony on May 22, 2026 that was set to formalize government authority to test frontier AI models before their public release. The cancellation came hours before the scheduled signing — some invited CEOs had already boarded planes when they learned the trip was pointless.
Key takeaways
- Trump canceled the EO signing on the day of the ceremony, hours before the scheduled time
- Elon Musk and Mark Zuckerberg reportedly urged Trump to scrap the executive order
- OpenAI publicly supported the signing
- The AI industry lobbied against the EO, fearing 90-day testing windows would delay model releases
- Trump cited concern that the order could be an innovation "blocker" in the AI race with China
What Was in the Order
According to the New York Times, the EO was designed to let the US government identify security vulnerabilities in frontier AI models before release — particularly to protect banks, power grids, and other critical infrastructure from AI-enabled cyberattacks. The administration also wanted to expand the number of AI labs voluntarily submitting to government vetting, a practice that had so far been sporadic.
The push came after Anthropic flagged serious cybersecurity risks with its Mythos model in an internal report, alarming members of the administration. The Commerce Department and the Office of Science and Technology Policy began recommending systematic testing, and a draft EO landed on the President's desk.
Musk, Zuckerberg, and Industry Lobbying
Semafor reported that xAI founder Elon Musk and Meta CEO Mark Zuckerberg actively pressured Trump to cancel the ceremony. Both reportedly argued that testing requirements could slow model development and hamper competition with Chinese AI. The broader tech industry lobbied collectively — a key point of friction was the proposed timeline: the administration wanted to test models up to 90 days before release, while AI labs pushed for a maximum of 14 days.
Musk denied involvement on X, writing "this is false" and claiming he did not know what was in the EO.
OpenAI — according to Semafor — publicly "supported" the signing, placing it in a distinctly different position from Musk and Zuckerberg.
Former AI advisor David Sacks — whose special government employee designation expired in March 2026 — also joined the push to delay the signing, Semafor reported. The Information noted that Sacks' departure from his formal role created a 'leadership vacuum' in the White House's AI structure, though he continues to visit weekly.
Trump: 'I Didn't Want Blockers'
Trump told reporters he decided against the order because he "didn't like certain aspects of it." He elaborated by invoking the China competition: "I really thought [the order] could have been a blocker. I think it gets in the way of — you know, 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."
"I really thought [the order] could have been a blocker. I think it gets in the way of — you know, 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." — Donald Trump, President of the United States, May 22, 2026.
Vice President JD Vance, who had said the day before that the administration was prioritizing "protecting people's data" and "people's privacy" after concerns about Mythos were raised, apparently did not have the final say.
Context: US vs China in the AI Governance Race
The irony of the situation is that China — often depicted as a country without AI regulation — is actually accelerating on governance. In April 2026, Beijing issued new rules requiring domestic AI firms to establish internal "AI ethics review committees." In May, the State Council outlined a 2026 legislative work plan covering "comprehensive legislation for the sound development of AI." The US, by contrast, has no federal AI regulations.
Lizzi C. Lee of the Asia Society Policy Institute told the South China Morning Post that Trump and Beijing face an analogous dilemma: how to guard national security interests without slowing frontier development. Lee suggests that a narrowly focused definition of testing — targeting only national security risks — probably would not slow leading US AI labs significantly.
Why This Matters
The canceled signing is not just a political episode — it is a signal moment in the US debate over AI governance. For the first time in years, the federal administration was close to a concrete oversight mechanism for frontier models, and pressure from a handful of CEOs was enough to block the initiative at the last minute.
The pattern is uncomfortably familiar: the White House proposes, Big Tech lobbies, the proposal is shelved or watered down. But the stakes with AI are higher than in earlier battles over regulating social media platforms. Frontier models can directly affect critical infrastructure, financial systems, and cybersecurity — which was precisely the rationale for the EO.
There is also a tension within the administration itself. National security-focused agencies know the time for governance is now — but they lack the political leverage to override the Commerce Department and the tech-friendly OSTP.
What's Next?
- Trump has not announced a timeline for re-signing the EO or specified what changes would be required — the document's status remains uncertain
- The US and China agreed at the Xi summit to launch an "intergovernmental dialogue on AI" — the first working meeting in this format will test whether governments can cooperate on AI safety without regulating their own firms
- Several US states are considering their own AI regulations — without federal action, pressure for state-level legislation may increase
