The U.S. government and the artificial intelligence industry are discussing a framework that would streamline market releases for domestic models, provided their capabilities do not exceed those of leading open-source Chinese models. Under the proposed guidance tied to the June artificial intelligence executive order, models at or below this capability ceiling would bypass mandatory government vetting before public release.

The executive order specifically targets cybersecurity risks. It calls for development of a classified benchmarking process to assess these capabilities, which would establish the threshold for the new safe harbor. Models exceeding the Chinese capability floor would remain subject to the existing framework, including its voluntary approval mechanisms.