White House Keeps Its New AI Testing Rules Behind Closed Doors

White House Keeps Its New AI Testing Rules Behind Closed Doors
The White House has drawn up new rules for testing America’s most powerful AI models—then kept the rulebook largely out of public view. The result is a national-security framework built around cooperation with industry, but shadowed by questions over who gets to see it and who gets left out.

On Monday, the administration said it had met the deadline in its June executive order to create a voluntary model-evaluation framework. A White House official said, “The voluntary framework outlined in the June 2nd executive order was complete by the deadline,” while adding that discussions with industry on implementation were underway.

The proposed process is designed to let developers determine whether models still under development fall within its scope. If they do, the government could receive access for as long as 30 days before release, under restrictions covering confidentiality, cybersecurity, insider threats, intellectual property and nondisclosure. The administration argues that discretion is warranted: “Just because things are unclassified that doesn’t mean we are going to broadcast them to everyone,” the official said.

On Tuesday, the White House held staff-level briefings with industry, with Anthropic, OpenAI and Google expected to attend. That meeting sharpened the divide between the companies inside the room and those outside it. The administration has not committed to publishing the framework, leaving other firms, researchers, policymakers and allies to infer how a policy with global security implications will work.

People briefed on the meetings say the framework focuses on closed-source frontier models with state-of-the-art capabilities and potential national-security risks. Open models are excluded, and the document says it should not be read as restricting them after release. Yet the key thresholds—what counts as “state-of-the-art,” or a national-security risk—remain unclear.

For the White House, secrecy is a safeguard for sensitive testing. For critics and excluded companies, it is the central flaw: a voluntary system meant to govern frontier AI is taking shape without a public map of its boundaries.

Continue reading https://foxvector.com/stories/019fcf05-65ef-3106-7045-1f9fee85a8be