Trump Administration Shifts Toward Federal AI Model Safety Reviews

The Donald Trump administration is pivoting from a laissez-faire approach toward formal government oversight of artificial intelligence. The White House is considering an executive order to establish a vetting system that would require next-generation AI models to undergo federal safety reviews before public release, a process National Economic Council Director Kevin Hassett compared to FDA drug approvals.

This policy shift was largely prompted by the development of Anthropic's Mythos model, which demonstrated an exceptional ability to identify thousands of cybersecurity vulnerabilities in operating systems and browsers. The administration is now forming an AI working group comprising government officials and industry leaders from Google, OpenAI, and Anthropic to develop evaluation protocols, potentially mirroring a safety-standard model in the United Kingdom.

In early May 2026, the Department of Commerce's Center for AI Standards and Innovation (CAISI) expanded its voluntary testing program. Microsoft, Google DeepMind, and xAI agreed to provide early access to unreleased models for pre-deployment security testing. Simultaneously, the Department of Defense signed agreements with seven tech firms, including Amazon and SpaceX, to deploy AI tools on classified military networks.

These developments follow a period of intense friction between the administration and Anthropic. The Department of Defense previously designated Anthropic as a supply-chain risk and terminated federal relationships after the company refused to remove safeguards against the use of AI for autonomous weapons and mass surveillance. While Vice President JD Vance and other officials continue to warn against excessive regulation, the administration is now prioritizing the mitigation of AI-enabled cyberattacks to protect national security.