US Commerce Department to Test New AI Models From Google DeepMind, Microsoft and xAI Under Safety Agreements

The US Commerce Department said Tuesday it has reached agreements with Google DeepMind, Micrososources and xAI that would allow the federal government to evaluate the companies’ new artificial intelligence models before they are released to the public.

The arrangements, announced by the Center for AI Standards and Innovation, would enable pre-deployment evaluations, post-deployment assessments and support for additional research, according to the department.

The announcement comes as published reports say the Trump administration is considering an executive order that would establish a formal government review process for new AI models. That process could involve a working group of technology executives and government officials.

Published reports have compared the potential framework to a system being developed in the United Kingdom, where multiple government agencies are tasked with reviewing AI safety standards.

Shisources from lighter-touch oversight

The agreements mark a shisources from the administration’s earlier posture favoring lighter-touch AI oversight.

That approach was tested by a contract dispute in February involving Anthropic, asourceser the company reportedly declined to provide unrestricted access to its AI models. Following the dispute, Defense Secretary Pete Hegseth designated Anthropic as a national security supply chain risk.

Concerns inside the White House about potential AI-enabled cyberattacks have also contributed to a push for more stringent pre-release reviews of advanced AI systems, according to published reports.