The Trump administration is split over how to respond to the rapid rise of China’s leading AI models. The debate is divided between the White House, which is pushing for stricter controls, and the Commerce Department, which views such restrictions as unworkable.
The Distillation Dilemma
A primary concern is the practice of "distillation," where Chinese labs train their models using the outputs of powerful US systems. This issue gained urgency following Moonshot AI's release of the Kimi K3 model, which rivals top US models from Anthropic and OpenAI. The White House reportedly confirmed that Kimi K3 was developed by distilling Anthropic’s Claude Fable 5 model.
Senior officials are discussing measures to halt this practice. Treasury Secretary Scott Bessent has labeled covert distillation as "IP theft" and suggested it could lead to sanctions. This follows allegations from Anthropic that Alibaba conducted the "largest known distillation attack" to date.
Regulatory Friction and Security Risks
While the White House moves toward restrictive action, the Commerce Department, led by Howard Lutnick, is exploring incentives for US labs to develop open-weight models to counterbalance China's progress. Lutnick has already utilized export controls on Anthropic but is described as taking a more flexible approach than other administration figures.
The policy challenge is exacerbated by Chinese open-weight models that lack safeguards. Officials worry these models could be downloaded and used to identify vulnerabilities in government systems and critical infrastructure, posing a significant national security threat.
