White House pauses public AI model testing by federal standards unit

Published June 9, 2026

Score

Why it matters

The White House has halted public assessments by the Center for AI Standards and Innovation (CAISI), a federal unit responsible for testing advanced artificial intelligence models. National Cyber Director Sean Cairncross and other administration officials cited national-security concerns as the reason for stopping the publication of evaluation reports. The decision reverses CAISI's recent public-facing role, which had earned approval from AI developers for its transparent testing methodology.

The halt follows the June 2, 2026 White House AI order on "Promoting Advanced Artificial Intelligence Innovation and Security," which prioritizes classified benchmarking and coordination with the NSA and CISA over public disclosure. The scope of CAISI's future work—whether the unit will continue operating under classified protocols, shift to advisory functions only, or face restructuring—remains unclear.

Attorneys tracking AI regulation should monitor how this shift affects the administration's stated commitment to balancing innovation with security oversight. The move signals a tightening of transparency around frontier AI capabilities at a moment when more powerful models are entering the market. This could reshape disclosure obligations for companies developing advanced systems and may influence how federal agencies approach voluntary model-access agreements with private developers.