US Administration Divided Over Response to Chinese AI Advancements

The Trump administration is experiencing a significant internal policy split regarding the regulation of Chinese artificial intelligence. According to multiple people familiar with the deliberations, the White House is advocating for stricter controls on Chinese AI systems, while the Commerce Department views such restrictions as unworkable. This debate has emerged as a flashpoint for the administration, particularly following the release of Moonshot AI’s Kimi K3 model, which rivals the capabilities of top US systems from Anthropic and OpenAI.

Senior White House officials have discussed measures to prevent Chinese labs from developing their models by training them on US models, a practice known as distillation. This focus on distillation has intensified after Anthropic alleged that Alibaba carried out the largest known distillation attack against it to date. The White House announced that Moonshot had developed Kimi K3 by distilling Anthropic’s Claude Fable 5 model, a move that has heightened concerns over intellectual property theft.

Treasury Secretary Scott Bessent has described covert distillation as IP theft and suggested it could lead to sanctions. While the White House is expected to take some form of presidential action to address these issues, it is not anticipated to take the form of an executive order. A White House spokesperson declined to comment on the reporting.

Conversely, the Commerce Department, led by Howard Lutnick, is exploring alternative strategies. Lutnick has spoken with leaders at several AI labs and is considering ways to incentivise the creation of US open-weight models to counterbalance Chinese advancements. The department oversees export controls through the Bureau of Industry and Security, and Lutnick has previously imposed export controls on Anthropic, though he is described as more freewheeling than other officials.

Despite proposals circulating within the White House, no formal request for input has been transmitted to the Commerce Department yet. The administration is increasingly concerned about open-weight models from China, which lack safeguards against hacking government systems and can be downloaded by anyone. This follows earlier policy headaches regarding Anthropic’s Claude Mythos and Fable 5 models, which were noted for their ability to find vulnerabilities in critical infrastructure.