Trump Signs AI Executive Order Granting Federal Government Early Access to ‘Frontier’ Models
Table of Contents
President Trump has signed an executive order aimed at tightening the intersection of national security and artificial intelligence, granting federal agencies a window of early access to the world’s most powerful AI models before they hit the general market. While framed as a cybersecurity necessity, the order has triggered immediate alarm among policy experts who argue it grants the administration dangerous levels of discretion over which companies thrive in the AI ecosystem.
The 30-Day Pre-Release Window
The core of the order centers on what the administration calls “covered frontier models”—the most advanced, large-scale AI systems that possess capabilities potentially dangerous to national security. Under the new directive, the National Institute of Standards and Technology (NIST) will help establish a voluntary framework where the federal government receives access to these models for up to 30 days prior to their public or commercial release.
This is a significant pivot from a May draft of the order, which had proposed a much more restrictive 90-day review period. The reduction to 30 days suggests a compromise to avoid stifling the rapid release cycles of AI labs, but the underlying mechanism remains: the government now has a formal seat at the table during the final stages of model deployment.
The ‘Trusted Partner’ Problem
Beyond the review window, the order introduces a controversial element regarding distribution. It directs agencies to collaborate with AI companies to select “trusted partners” who will also receive early access to these models. In practice, this means the administration could influence the guest lists for exclusive preview programs, such as Anthropic’s Project Glasswing or similar early-access tiers for Claude and GPT-next iterations.
For policy analysts, this creates a precarious incentive structure. By controlling the flow of early access to the most capable tools, the executive branch could effectively “pick winners and losers” in the private sector. Juan Londoño, a policy analyst at the Cato Institute, noted that the lack of clear criteria for what constitutes a “covered frontier model” gives the administration immense discretion, warning that the system could be weaponized against companies that find themselves at odds with the White House.
Building a Cybersecurity ‘Clearinghouse’
To ground the order in operational utility, the directive tasks the Treasury Department with creating an “AI cybersecurity clearinghouse.” This entity is intended to act as a hub between the AI industry and critical infrastructure operators, coordinating how advanced AI is used for vulnerability scanning, discovery, and the subsequent patching of software flaws.
The order also expands the US Tech Force’s hiring pathways for Information Cybersecurity Specialists and steers federal grant money toward firms developing AI-driven vulnerability detection. These elements are generally viewed as pragmatic steps toward hardening US infrastructure against AI-assisted cyberattacks.
A Fragile Balance of Power
The tension within the order lies in its “voluntary” nature. While the framework is not mandatory, the pressure to comply is high for any AI lab seeking federal contracts or wanting to avoid more stringent, mandatory legislation. Samir Jain, VP of Policy at the Center for Democracy and Technology, praised the non-mandatory approach but echoed concerns that the “trusted partner” provision could become a tool for political punishment.
Former FTC chief technologist Neil Chilson argued that while any formal rule is an improvement over the previous informal approach, the current gaps in the order may allow short-term security concerns to override long-term economic growth and innovation. The consensus among critics is a call for Congress to step in and establish a codified legal framework to replace the broad discretionary powers granted by this executive action.
The White House has not yet provided further details on the specific benchmarks NIST will use to classify frontier models or the timeline for the clearinghouse’s launch.
