President Donald Trump abruptly canceled plans to sign a new executive order on artificial intelligence just hours before a scheduled White House ceremony. Trump cited concerns that the measure could potentially diminish America’s competitive advantage in AI technology, stating, “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 indicated dissatisfaction with the text of the proposed order, leading to the postponement of the Oval Office event with tech industry executives.

The aborted directive was intended to establish a governmental framework for vetting the national security risks associated with the most advanced AI systems prior to their public release. Described as a voluntary collaborative effort, the order sought participation from major U.S.-based tech companies, including Anthropic, OpenAI, and Google. This push for government oversight stemmed from escalating anxieties within the banking industry and other critical institutions regarding AI’s rapidly advancing capabilities to identify cybersecurity vulnerabilities in global software systems.

In April, Treasury Secretary Scott Bessent and then-outgoing Federal Reserve Chair Jerome Powell convened an urgent meeting with Wall Street CEOs at the Treasury Department. During this meeting, they specifically warned about the cybersecurity threats posed by Anthropic’s AI model, Claude Mythos. Bessent, speaking at CNBC’s “Invest in America Forum,” emphasized the model’s power and the necessity of ensuring banks were adequately prepared. This led to discussions among Republican allies about developing more effective methods for deploying AI tools into the hands of trusted cybersecurity experts.

Trump’s aspirations for the AI sector have frequently clashed with public fears about the technology’s broader societal impact on American life, job markets, and electricity consumption. While Trump had previously vowed to dismantle the AI safety regulations enacted by his predecessor, Joe Biden, and has consistently promoted the AI sector as a key driver for economic expansion, the Republican party itself remains divided on whether to fully embrace the industry or align with voters expressing skepticism. Furthermore, the government’s interest in collaborating with Anthropic on cybersecurity is complicated by an ongoing legal dispute with the company. In February, Trump had ordered all U.S. agencies to cease using Anthropic’s chatbot, Claude, following a public disagreement between the Pentagon and CEO Dario Amodei.

Serena Booth, a computer science professor at Brown University and former AI policy fellow, highlighted the visible internal conflicts within the administration regarding AI policy. She described the policy process as experiencing “whiplash” and “fractures,” exemplified by the public back-and-forth over the executive order. These divisions involve balancing the “reasonable idea” of testing highly capable AI models before release with concerns that overly lengthy government scrutiny could impede innovation and slow development. Booth acknowledged the significant potential cost to innovation and speed, recognizing the validity of both perspectives.

The White House has previously opposed state-level AI regulations, arguing they could hinder growth. Therefore, an executive order implying government screening of commercial AI models would have represented a notable shift in the administration’s strategy. Interestingly, similar screening mechanisms are already in motion; the Commerce Department had announced agreements earlier this month with Google, Microsoft, and Elon Musk’s xAI to evaluate their most powerful AI models, building upon previous agreements made by the Biden administration with Anthropic and OpenAI. However, the announcement later vanished from the Commerce Department’s website.

During a White House press briefing, Vice President JD Vance refrained from detailing the specific contents of the order but affirmed the administration’s commitment to fostering innovation while simultaneously addressing cybersecurity threats and data privacy. Vance articulated the president’s desire for the U.S. to lead the global AI race, emphasizing the need to protect citizens. He underscored the administration’s collaborative approach with tech companies to navigate the potential downsides of new, security-risky AI models, striving to balance safety with innovation. Dean Ball, a former White House tech policy adviser, characterized these disagreements as “healthy tension” within an administration historically cautious of regulating “frontier AI” companies. While acknowledging the political risks involved in such regulation, Ball expressed support for the administration taking the necessary time to refine its approach to AI policy.