Trump Signs Executive Order to Vet Top AI Models for National Security Risks

President Donald Trump signed an executive order on Tuesday establishing a new framework requiring developers of the most advanced artificial intelligence models to submit their systems for national security risk assessments before deployment. The order directs the Commerce Department's Bureau of Industry and Security (BIS) to create a mandatory vetting process for frontier AI models above specified compute thresholds.

The executive order, titled "Securing American Leadership in Artificial Intelligence," marks a significant escalation in the federal government's approach to AI governance. Unlike previous voluntary commitments sought by the Biden administration, the new order carries the force of regulation, requiring companies developing models trained on more than 10^26 floating-point operations to certify that their systems do not pose unacceptable risks to national security, critical infrastructure, or biosecurity.

Key Provisions and Industry Impact

Under the order, companies must provide the government with pre-deployment access to model weights, safety evaluation results, and red-teaming reports. A new AI Security Review Board — composed of officials from the Department of Defense, the intelligence community, and the Department of Homeland Security — will have 45 days to review submissions and can request modifications or impose deployment restrictions.

The order specifically highlights risks related to cyber offense capabilities, biological weapon design, and autonomous decision-making in critical systems. It also requires foreign companies seeking to deploy frontier models in the U.S. market to undergo the same review process, a provision widely interpreted as targeting Chinese AI labs including DeepSeek and ByteDance.

Industry Reaction

Reaction from major AI companies has been mixed. OpenAI, Google DeepMind, and Anthropic have generally expressed willingness to cooperate, noting that they already conduct extensive safety testing. However, smaller frontier labs and open-source advocates have raised concerns that the compute threshold could be lowered over time, potentially stifling innovation and open-weight model releases.

"We support the goal of national security vetting, but the executive order leaves too much discretion to agencies without sufficient transparency or appeal mechanisms." — Statement from a coalition of AI startups

The executive order is expected to face legal challenges from industry groups arguing that it exceeds executive authority. Congressional leaders have indicated they may introduce complementary legislation to provide a more durable statutory framework for AI safety oversight.

Source: White House fact sheet, Reuters, The Washington Post, and Bloomberg Government.