Trump Administration Revives Pre-Deployment AI Testing in National Security Pivot

The Trump administration has quietly reversed course on artificial intelligence oversight, signing agreements with three major technology firms to conduct pre-deployment safety testing of frontier AI models before public release.

The Commerce Department's Center for AI Standards and Innovation announced Tuesday it will evaluate systems from Google DeepMind, Microsoft, and xAI through what it calls "targeted research" and pre-deployment assessments. The agreements also renegotiate existing deals with OpenAI and Anthropic, first struck under the Biden administration in August 2024, to align with current White House priorities.

"Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris Fall said in a statement.

The shift marks a significant departure from the administration's initial posture emphasizing rapid innovation without regulatory constraints. Earlier policy documents prioritized beating China through accelerated development, with minimal government intervention in the deployment pipeline. The new testing regime suggests growing internal concern about security vulnerabilities, particularly following cybersecurity issues flagged in connection with Anthropic's Mythos model.

Reports indicate the White House is weighing broader executive action to formalize government review processes for new AI systems, potentially establishing a multi-agency working group with pre-clearance authority. Such a structure would resemble oversight frameworks operating in the United Kingdom, where layered reviews confirm safety standards before market entry.

(The policy evolution comes as David Sacks, who initially served as White House AI and Crypto Czar before transitioning to co-chair the President's Council of Advisors on Science and Technology in March, has publicly defended the administration's approach against polling showing three-quarters of Americans believe the government is doing too little to regulate AI.)

The Commerce Department's AI Safety Institute, established by Biden executive order in 2023 and renamed under Trump, was initially expected to pivot from safety evaluation toward acceleration mandates. Instead, the institute has continued publishing technical assessments, including an evaluation of China's DeepSeek model, and soliciting public comment on secure AI agent deployment.

The voluntary testing agreements closely mirror the Biden-era framework, raising questions about whether the administration's rhetorical emphasis on deregulation will translate into substantively different oversight. Industry observers note that while the deals remain non-binding, they provide the government early visibility into capabilities that could pose national security risks, from advanced cyber-offensive tools to dual-use applications in weapons systems.

The timing coincides with the Department of Defense finalizing agreements with seven technology companies to integrate AI tools into military decision-making, underscoring the strategic importance Washington places on maintaining technological superiority even as it grapples with governance challenges.