The Trump administration on Tuesday told leading AI companies it will not conduct safety testing on open-weight AI models, according to Reuters citing two people familiar with the matter. According to ChainCatcher, the move marks a shift in the U.S. government's approach to AI regulation.
Open-weight models allow developers and users to download, modify, and deploy model parameters freely without relying on closed API access. Before this decision, the U.S. AI Safety Institute and other agencies had conducted safety evaluations of frontier models. The White House and relevant departments had not publicly commented.