White House Hides AI Rules While Exempting Open Models

White House Hides AI Rules While Exempting Open Models
The White House has finished its new AI testing framework, but the central rules governing the country’s most powerful models are staying behind closed doors. The result is a voluntary safety regime that promises scrutiny for some frontier systems while leaving much of the AI world guessing.

The process began with President Trump’s June executive order, which asked developers to share certain advanced models with the federal government before release to assess cyber risks. The order made the model-benchmarking process classified, but did not require the broader voluntary framework itself to be secret.

By Monday, the administration said it had met its deadline. “The voluntary framework outlined in the June 2nd executive order was complete by the deadline,” a White House official said, adding that industry discussions on next steps were underway. The administration’s position is that confidentiality need not be synonymous with public disclosure: “Just because things are unclassified that doesn’t mean we are going to broadcast them to everyone.”

On Tuesday, officials briefed industry participants, including companies such as Anthropic, OpenAI and Google. But firms not invited, as well as policymakers, researchers and U.S. allies, were left without the framework’s text or a clear view of who counts as a trusted partner. The White House declined to comment on the decision not to publish it.

What has emerged points to a narrow system: a covered frontier model is reportedly closed-source, state-of-the-art and tied to national-security risks, with government access possible for up to 30 days before release. Yet neither “state-of-the-art” nor “national security risk” is clearly defined, creating uncertainty particularly for smaller providers trying to anticipate Washington’s expectations.

Open models are explicitly excluded, and the framework says it cannot be used to restrict them after release. Yann LeCun amplified that development in a repost describing the exemption of open models from pre-release frontier-capability testing. Supporters of openness may see a safeguard; critics see a split regime whose most consequential details remain private.

Continue reading https://foxvector.com/stories/019fd2e2-0500-3784-705e-25e1c4a5f5df