Trump Pulls AI Executive Order Hours Before Signing Ceremony

President Trump abruptly delayed the signing of a landmark executive order on artificial intelligence Thursday afternoon, telling reporters that he had pulled the order at the last minute because it could interfere with American competitiveness in AI. The cancellation came hours before a planned Oval Office ceremony at which the CEOs of OpenAI, Anthropic, and other major labs were expected to appear.

The decision throws into question how the administration plans to handle AI oversight after weeks of negotiation between the White House, the National Cyber Director's office, and industry. It also leaves an open lane for the next move on AI policy, whether that comes from Congress, state attorneys general, or a revised order down the road.

What the order would have done

The draft executive order, leaked in pieces over the past week, would have established a voluntary framework under which leading AI labs share new frontier models with the federal government at least 90 days before public release. The National Security Agency was expected to handle classified testing of those models, and the Treasury Department would have helped run a "clearinghouse" to surface and fix cybersecurity flaws.

The framework was voluntary, but the political signal would have been substantial. By formalizing pre-release access, the White House would have effectively codified the kind of cooperation that some labs have offered ad hoc since 2024. It would also have given national security agencies a structured seat at the table for the first time.

Why this matters: voluntary or not, an executive order tends to set the terms of future debate. Congress writes laws against the backdrop of what the executive branch already does. State regulators take cues from federal positioning. The absence of an order leaves the field unstructured.

What Trump said

Speaking to reporters, Trump said: "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." He added that "I really thought it could have been a blocker," referring to the draft text. The president did not commit to a new signing date or signal what would change in a revised draft.

The language is significant. Trump has used the same framing for tariffs, export controls, and chip policy: anything that slows the American AI industry relative to China is off the table. By pulling the order on that rationale, the White House is telling labs that the bar for new oversight is whatever they themselves consider non-disruptive.

Why the White House paused

Reporting earlier this week described internal friction over the order's scope. The voluntary 90-day pre-release window was one sticking point. Some advisors argued it would slow product releases and give competitors abroad a head start. Others, particularly on the cyber side, argued the window was necessary because frontier models can now exploit vulnerabilities faster than traditional security tooling can patch them.

The trigger for new urgency was Anthropic's Mythos model, which the company demonstrated could autonomously chain together cybersecurity exploits at speeds the defensive industry has not matched. That demonstration moved the White House from a hands-off posture into drafting an order in a matter of weeks. The same posture is what some inside the administration now think went too far, too fast.

Industry reaction

Labs that had publicly supported voluntary pre-release access spent Thursday in a holding pattern. None of the major companies released formal statements immediately after the cancellation, in part because the executive branch had not given them a clear signal on whether the order is dead, delayed, or being rewritten.

Privately, several executives told reporters they had been preparing for the signing as if it would happen. The companies had reportedly already aligned legal teams around the voluntary terms and were ready to publicly endorse the framework. That work is now on ice.

OpenAI, Anthropic, and Reflection AI had each been part of the drafting conversations, according to earlier Axios reporting. Their CEOs were expected at the signing. With the event canceled, the CEOs were not present at the White House Thursday afternoon.

What this means for AI oversight

The immediate impact is procedural: there is no new federal AI oversight framework as of Thursday evening. The longer-term impact depends on how Congress and the states respond. Connecticut's AI Responsibility and Transparency Act, signed earlier this year, demonstrates that state legislatures will move if the federal level stays quiet. California, New York, and Texas all have draft bills in committee.

For labs, the cancellation removes the cleanest path to formalizing the safety claims they make in public. Voluntary pre-release access was supposed to be a credible signal — proof that the companies were willing to be checked. Without an order, those claims revert to being marketing statements.

For national security agencies, the loss is more direct. The order would have given the NSA and Treasury Department structured access to unreleased models. That access is the only realistic way for the government to maintain situational awareness about what's coming out of frontier labs. Without the order, those agencies remain dependent on whatever the companies voluntarily share.

What's next

Trump did not commit to a revised draft or a new timeline. White House staff have not said publicly whether the order will be reworked or shelved. The most likely outcome, based on reporting from inside the administration, is a slimmed-down version that drops the 90-day window in favor of something narrower — perhaps an information-sharing arrangement for specific national security use cases only.

Tech CEOs who had been working with the White House on the order will likely use the next several weeks to lobby for terms they consider compatible with shipping speed. State legislators, meanwhile, may move faster on their own bills, betting that federal inaction creates a vacuum they can fill.

The bottom line

A voluntary pre-release oversight framework was on the table this morning. By the end of the day, it was off. The reasoning — that any new oversight could slow US competitiveness — sets the terms for what comes next. The next time someone in the White House proposes formal AI oversight, the burden will be on them to prove it doesn't cost the United States its lead. That is a hard bar to clear.