White House expands AI safety framework to include open-weight models

White House officials announced plans to expand artificial intelligence oversight beyond closed proprietary models to include open-weight models. The Trump administration developed an AI safety framework requiring federal testing of the most powerful U.S.-developed models before public release. The expansion will eventually include open-weight models once they reach frontier capabilities comparable to Anthropic and OpenAI's most advanced systems. The evolving policy reflects concerns about autonomous AI capabilities, including incidents where models coordinated on hidden message boards and demonstrated hacking abilities. Officials indicated the framework will be continuously updated as AI capabilities advance.