US AI Rulebook Kept Private, Angering Smaller Labs

Major artificial intelligence companies met with U.S. government officials at the White House on Tuesday to discuss a new voluntary framework for reviewing "frontier AI models." Participants in the meeting included leading organizations such as OpenAI, Anthropic, Google, Meta, and Nvidia. The outcome of this high-level discussion was an agreement on a proposed arrangement that will allow the government to review these advanced AI models. However, this framework, which has been under development for the past two months following an order from President Trump in June, will not be made public. The administration has indicated no plans to publish the details of this agreement.

Significant questions remain unanswered for the public and a substantial portion of the AI industry regarding the specifics of this framework. Key details such as which AI models will be subject to review, the specific thresholds that define a "covered frontier model," and the identity of the "trusted partners" who will receive early access to the world's most powerful AI technologies are currently unknown. A directive issued in June already indicated that the benchmarking process used to classify these "covered frontier models" could be classified, with the final determination resting with the director of the National Security Agency (NSA). This secrecy has fueled concerns among smaller AI developers and those working with open-source models.

A source familiar with the proceedings reported that only a limited number of companies were present during the briefing where the framework was discussed. This exclusivity highlights a broader apprehension within the AI community: that the development of U.S. AI regulation is increasingly being influenced by private discussions among a select group of dominant firms. Smaller laboratories and open-source AI initiatives fear that these safety frameworks could inadvertently shape the broader market landscape. Their concern is that such frameworks might favor and entrench the use of closed-source frontier models that have received implicit or explicit "government approval," potentially creating barriers for innovation and competition from less established players.

The lack of transparency surrounding the framework's contents and the review process raises questions about fairness and accessibility in the rapidly evolving field of artificial intelligence. While major corporations have agreed to the voluntary framework, the broader AI ecosystem is left in the dark about the rules governing the development and deployment of the most advanced AI systems. This situation underscores a growing tension between the need for robust AI safety measures and the desire for open innovation and equitable market access in the AI industry.