An anticipated executive order on artificial intelligence from Donald Trump was abruptly shelved after intense eleventh-hour pressure from Silicon Valley leaders, including Elon Musk and Mark Zuckerberg, according to people familiar with the matter. The decision, which came just hours before a planned signing ceremony, underscored how deeply influential the tech industry remains in shaping federal policy on artificial intelligence, even at the highest levels of government.
The draft order, developed over weeks of internal White House discussions and reviewed by industry stakeholders, would have introduced a voluntary federal vetting system for advanced AI models. As described in materials cited by The Washington Post, the framework would have allowed government agencies to evaluate frontier systems up to 90 days before public release, with the goal of identifying security vulnerabilities and testing for potentially harmful capabilities. Officials framed it as a balance between innovation and safety, emphasizing that participation would not involve mandatory licensing or preclearance requirements.
However, the proposal quickly became controversial within both political and industry circles. Critics inside the administration and Silicon Valley warned that even a voluntary system could evolve into a de facto regulatory gatekeeper. That concern was central to the arguments made in last-minute conversations with the president, which included participation from former AI and crypto adviser David Sacks. According to multiple people familiar with the discussions, Sacks and other executives cautioned that the policy could slow product rollouts and stifle competition at a critical moment in the global artificial intelligence race.
The turning point came after a series of private calls between Trump and leading tech figures. These conversations reportedly highlighted fears that the United States could lose strategic ground to China if companies were required to share model details with regulators before release. Some administration officials had argued the opposite, warning that without structured oversight, adversaries could exploit weaknesses in advanced systems to launch cyberattacks or manipulate emerging technologies. The disagreement reflected a broader divide inside the administration over how aggressively to regulate artificial intelligence while still maintaining American technological leadership.
The political stakes surrounding the decision were unusually high. The White House had already sent invitations to executives across the tech sector for a formal signing ceremony scheduled for Thursday afternoon, signaling confidence that the order would proceed. Some industry leaders were even en route to Washington when the event was abruptly canceled. Shortly before the scheduled announcement, Donald Trump told reporters inside the Oval Office that he had decided against signing the order, explaining that he “didn’t like” the final draft and believed it could have created unnecessary barriers for innovation.
“I really thought that could have been a blocker,” Trump said, according to remarks reported by The Washington Post, adding that while he recognized concerns about safety, he wanted to ensure that artificial intelligence development in the United States would not be slowed. His comments highlighted a central tension in the policy debate: how to encourage rapid technological advancement while still addressing risks tied to increasingly powerful AI systems capable of writing code, identifying vulnerabilities, and potentially being weaponized.
Inside the administration, reactions to the decision were mixed. Some officials expressed surprise that the order had been pulled so late in the process, especially after weeks of negotiation and refinement. Others acknowledged that the influence of Silicon Valley leaders, including Musk and Sacks, had been decisive. White House staff involved in drafting the order, including science adviser Michael Kratsios, Staff Secretary Will Scharf, and National Cyber Director Sean Cairncross, had previously briefed Sacks on the proposal. At the time, he reportedly indicated he could accept the framework, though he pushed for narrower timelines and reduced regulatory friction.
The proposed system itself had been designed to avoid heavy-handed regulation. Companies would have been asked to voluntarily submit frontier AI models ahead of release, allowing government experts to test them for dangerous behaviors or security flaws. The aim, according to the draft cited by The Washington Post, was to create a collaborative safety pipeline without imposing legal barriers to deployment. Yet even this softer approach became a point of contention, as industry leaders warned that voluntary participation could gradually become expected practice, effectively transforming into mandatory compliance under future administrations.
Tensions also reflected deeper political dynamics within the coalition supporting Trump’s return to power. Figures such as Musk and Sacks are not only influential voices in technology policy but also significant political donors whose priorities intersect with Republican economic strategy. At the same time, concerns are growing within parts of the political base about artificial intelligence displacing jobs, raising energy costs, and concentrating power in the hands of a few dominant companies. These competing pressures have made AI governance one of the most volatile policy arenas in Washington.
Another layer of complexity comes from rapid advances in AI capabilities themselves. Systems developed by companies like Anthropic and other frontier labs have demonstrated increasingly sophisticated reasoning and, in some cases, the ability to identify security weaknesses in software systems. These developments have intensified debates over whether proactive government testing is necessary to prevent misuse or whether such oversight risks slowing down innovation in a strategically critical sector.
