Trump's AI Order Hands Classified Frontier Model Review to NSA

When the White House released Executive Order 14409 on June 2, the framing was cybersecurity: hardening federal systems against AI-enabled threats and creating a clearinghouse for sharing vulnerability information. The accountability story, as TechPolicy.Press reports, cuts differently. The NSA, acting through the Secretary of Defense, is delegated to build the benchmarking process that determines when a model earns "covered frontier model" designation, and that benchmarking process will be classified, with the NSA Director making the final call.

The practical consequence is a gap with no obvious precedent in U.S. product safety regulation. There is no requirement for unclassified summary reporting to Congress, no role for the Government Accountability Office, and no ability to acquire information through freedom of information requests, neither the benchmarking criteria nor the results. Smaller AI companies, academic researchers, journalists, and civil society organizations who study AI risk will have no way to know what the threshold is, how it was set, or how consistently it is applied. Companies whose models are evaluated will presumably learn how they scored; the public will not.

The piece draws a pointed contrast to how the government handles other consequential product safety calls. Drug approvals, vehicle safety evaluations, and environmental permitting all involve some combination of published criteria, notice-and-comment processes, and records eventually reviewable by courts, journalists, or independent experts. The EO places the most consequential judgments about powerful AI systems almost entirely inside the classified world of an intelligence agency whose core mission and culture are built around secrecy.

The order's scope exclusions also matter. The piece notes the order focuses only on cyber risks, leaving out any mention of privacy, civil liberties, or public safety. By framing AI and data almost entirely as a cybercrime issue, the analysis argues, the order sidesteps the implications for surveillance, data brokerage, and behavioral profiling by both government and corporations.

What the reporting does not settle is whether Congress has appetite to impose unclassified reporting requirements, or whether allied governments are actively moving toward alternative frameworks in response. The accountability question is structural: when evaluations of the most capable AI systems happen inside a classified intelligence-agency process, the oversight mechanisms that exist in other safety-critical domains (GAO review, judicial scrutiny, independent research access) are not simply delayed, they are structurally absent.