WASHINGTON — President Donald Trump signed an executive order May 21 creating a voluntary framework for federal vetting of top artificial intelligence models, focusing on AI systems deemed critical to national security.
The White House characterized the directive as a voluntary collaboration with U.S.-based tech companies, including Anthropic, OpenAI and Google — sometimes called "frontier labs" for their work on advanced AI systems.
Under the order, frontier labs can share cutting-edge cyber models with the government. The goal is to secure critical infrastructure and strengthen federal cyber defenses. Several company executives had planned to attend a signing ceremony, but Trump signed the order without one.
The administration said the approach avoids excessive government intervention. The White House said it is not conducting oversight of all new models because such broad regulation would have chilling effects on free speech and innovation.
Experts at the Council on Foreign Relations said the "covered frontier model" designation extends beyond a single model to encompass an entire AI system — including integrated models, data pipelines and deployment architecture.
Those systems are relevant to national security in specific domains: autonomous operations, cyber operations and scientific reasoning, where state-of-the-art AI performance is critical.
Proponents argue the voluntary framework protects American companies from heavy regulation and lets them outpace international competitors in AI development.
Critics have questioned whether a non-mandatory approach leaves oversight gaps. The administration frames the voluntary collaboration as a national security measure that limits tech monopolies and advances U.S. technological leadership — though its effectiveness will depend on how many key AI developers choose to participate.
