Michelle De Mooy writes in Tech Policy Press that the U.S. government’s secretive framework for reviewing frontier AI models before release, created under Executive Order 14409, lacks basic transparency safeguards. She proposes five standards the framework should meet, including publishing review criteria, setting default release time limits, disclosing which organizations get pre-release access, and ensuring due process for developers. De Mooy argues that legitimate regulatory authority is defined not by its scope but by how transparently it is exercised, warning that without these safeguards the framework amounts to unchecked executive discretion.
