Washington is keeping its AI rulebook private. Smaller AI labs aren’t happy.

The White House convened a meeting with major AI companies to address the confusion surrounding the regulation of frontier AI models. OpenAI, Anthropic, Google, Meta, Nvidia, and other leading AI firms participated in the discussion. However, the proposed voluntary framework, aimed at allowing the government to review these models, will remain private, with the general public and much of the industry unaware of its contents.

The framework's specifics, such as the models covered, thresholds, and "trusted partners" with early access, are still unclear. Smaller and open-source AI labs express concern that such safety frameworks could influence the market structure and favor closed-source models approved by the government. The review process involves submitting models for up to 30 days before release, but the criteria for classification as a "covered frontier model" and "national-security risk" are not explicitly defined.

Open-weight models have been excluded from the framework, potentially benefiting companies and leaving out the AI models that the U.S. is most worried about. The framework also mandates high-security environments for the submitted models, restricting access to company staff. Critics argue that the lack of transparency may result in a decision-making process that is arbitrary and burdensome, rather than ensuring public safety.