The administration considers imposing restrictions on Chinese open-weight AI models to mitigate national security risks
The administration is currently evaluating potential restrictions on Chinese open-weight AI models, such as the Kimi K3 and DeepSeek-R1, to address growing national security and economic concerns. While these models offer sophisticated coding capabilities and low costs, they often lack sufficient guardrails and exhibit deep ideological alignment with the Chinese Communist Party. For instance, the DeepSeek-R1 model has shown a susceptibility to security vulnerabilities like prompt injection, and its ideological alignment often presents Beijing's version of history as the global default. To address these risks, the administration announced a need for robust testing and public evaluation of Chinese models to rectify the current imbalance in transparency. By establishing clear testing standards, the U.S. government can provide a basis for evidence-based restrictions that are harder for Beijing to dismiss as protectionism. Furthermore, the administration is considering using the Information and Communications Technology and Services (ICTS) program to designate specific developers like Moonshot on the basis of ties to Chinese security services. The goal is to maintain America's edge by proactively degrading China's AI development through stronger export controls and closer coordination among American developers.