The Trump administration is preparing a formal regulatory mechanism that would route frontier artificial intelligence (AI) models through federal safety evaluations before they are released to the public, according to senior White House officials. The shift, building on a series of bilateral agreements struck earlier this month with major AI developers, represents the administration's first significant move from a hands-off posture toward direct oversight of cutting-edge model deployment.
A proposed FDA-style review
National Economic Council Director Kevin Hassett confirmed this week that the White House is weighing an executive order to formalize the process. <cite index="2-5,2-6">Hassett, speaking in a Fox Business interview, said the administration is considering an executive order to ensure new AI models are secure before they are released publicly, and compared the approach to how the Food and Drug Administration (FDA) evaluates drugs for safety.</cite> <cite index="2-7">He said officials are studying an order that would give a clear road map for how future AI systems that could create vulnerabilities would go through a process so they are released 'in the wild' only after being proven safe, similar to an FDA drug.</cite>
A draft of the order circulating among agencies is reportedly split into two tracks. <cite index="6-1,6-2,6-3">One section, titled 'covered frontier models,' defines which AI systems would be eligible for a voluntary framework allowing early government review and up to 90 days of pre-public access for federal evaluators. A second section outlines a voluntary cybersecurity 'clearinghouse' formed by the Treasury Department, other agencies and AI companies to identify and fix security vulnerabilities in unreleased models.</cite>
CAISI agreements with major labs
The policy push follows action by the Department of Commerce's Center for AI Standards and Innovation (CAISI), the successor body to the Biden-era US AI Safety Institute. <cite index="5-4,5-5">CAISI announced agreements with Google DeepMind, Microsoft and Elon Musk's xAI that will allow the US government to evaluate AI models before they are publicly available, with the center conducting pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security.</cite>
<cite index="5-6,5-7">The announcement builds on CAISI's previous partnerships with OpenAI and Anthropic from 2024, which have been renegotiated to reflect CAISI's directives from Commerce Secretary Howard Lutnick and America's AI Action Plan.</cite> <cite index="2-10">CAISI has conducted 40 evaluations to date, including on some models that have yet to be released.</cite>
A reversal in tone
The pivot is notable given the administration's earlier stance. <cite index="3-5,3-6">President Trump had positioned himself as a champion of unfettered AI development, rolling back Biden-era safety evaluation requirements and casting regulation as a threat to US competitiveness with China; the Biden administration's 2023 executive order had required AI developers to share safety test results with the government and directed federal agencies to set technology standards.</cite>
Industry observers attribute the change in posture in part to recent capability disclosures from frontier labs. <cite index="6-22">The administration had taken a more hands-off approach to AI regulation until recently, when Anthropic unveiled its Mythos model, which the company says can exploit cybersecurity vulnerabilities at an unprecedented pace.</cite> <cite index="3-8">The National Security Agency has reportedly gained access to Mythos and is conducting tests on it.</cite>
CAISI Director Chris Fall, who replaced former Anthropic researcher Collin Burns earlier this year, framed the program in measurement terms. <cite index="1-4,1-5">'Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,' Fall said, adding that the expanded industry collaborations help scale the work 'in the public interest at a critical moment.'</cite>
The executive order has not yet been signed, and the White House has declined to confirm a timeline.