The Trump administration’s planned AI safety framework is set to exclude Chinese “open-weight” AI models from certain U.S. government tests, according to separate reports citing people familiar with the matter and describing policy details. NDTV says the framework has not been publicly released and could remain private, and links the approach to an executive order signed in June that is intended to address AI security concerns.
Bloomberg reports that the White House has informed top U.S. AI firms that open-weight models developed by Chinese competitors will not be subject to government testing under the new framework. The accounts describe the decision as part of the administration’s broader implementation of the executive order, though they do not provide the full specifics of the testing regime, the criteria for exemption, or the extent of oversight that may still apply.
Both reports agree that the exemption concerns open-weight models from China and that the policy is tied to the June executive order and the related U.S. AI safety framework, which is still not clearly available in full form publicly.
