White House Divided Over Chinese AI Policy Response
Original: The White House Is Trying to Figure Out What to Do About Chinese AI
Why This Matters
US-China AI rivalry is now driving federal policy debates around export controls, IP enforcement, and open-weight model strategy.
The Trump administration is internally split on how to address China's rising AI capabilities. The White House favors stricter controls on Chinese AI models, while the Commerce Department views such restrictions as unworkable, following Moonshot AI's release of Kimi K3.
The Trump administration is engaged in an internal debate over how to respond to increasingly powerful Chinese AI models. The White House has pushed for stricter controls, while the Commerce Department—which oversees export controls through the Bureau of Industry and Security—has resisted, calling such measures unworkable. The immediate trigger is Moonshot AI's release of Kimi K3, which the White House alleges was developed by distilling Anthropic's Claude Fable 5 model—a practice of training one AI model using outputs from another. Senior officials are discussing presidential action to curb distillation attacks, though not via executive order. Treasury Secretary Scott Bessent called covert distillation 'IP theft' and warned it could lead to sanctions. The dispute was further inflamed after Anthropic alleged that Alibaba had conducted the 'largest known distillation attack' against it to date. Commerce Secretary Howard Lutnick has been exploring incentives to encourage US labs to develop open-weight models as a counterbalance to China's offerings. Earlier policy concerns centered on Anthropic's Claude Mythos and Fable 5 models and their potential to find vulnerabilities in government systems, but open-weight Chinese models—downloadable by anyone with minimal safeguards—have now become the administration's primary AI policy headache.