White House Grapples With Response to China's Fast-Advancing AI Capabilities

The Trump administration is internally divided on how to address China's rapidly improving artificial intelligence systems, as Beijing's AI labs release models that are beginning to match the capabilities of leading American competitors.

The Core Tension#

The disagreement pits hawkish elements within the White House, who want aggressive measures to block Chinese AI development, against the Commerce Department, which maintains that many proposed restrictions are not feasible to implement. The Commerce Department's influence stems partly from its oversight of export controls through the Bureau of Industry and Security.

The Distillation Problem#

A key flashpoint has been distillation attacks—the practice of training new AI models by extracting capabilities from existing, more powerful ones. White House officials have discussed strategies to prevent China's AI labs from using this technique to enhance their own systems. Anthropic recently alleged that Chinese tech company Alibaba carried out what it described as the largest known distillation attack against it. Treasury Secretary Scott Bessent characterized covert distillation as "IP theft" and suggested it could trigger sanctions.

The White House also publicly announced this week that Moonshot's recent Kimi K3 model—which rivals top US systems from Anthropic and OpenAI—was likely developed through distillation of Anthropic's Claude Fable 5 model.

Potential Presidential Response#

The administration is expected to take some form of presidential action, though sources indicate this is unlikely to be a formal executive order. However, no official requests for input have yet been sent to the Commerce Department, which is developing its own policy proposals.

Commerce leadership, led by officials including Lutnick, is exploring alternative strategies—such as creating incentives for US AI labs to develop their own open-weight models to compete with Chinese alternatives. Lutnick has reportedly held discussions with leaders at multiple AI companies in recent weeks, positioning himself between harder-line and more permissive regulatory approaches.

Broadening AI Concerns#

Earlier this month, the administration focused on concerns about Anthropic's models and their capacity to identify vulnerabilities in government and critical infrastructure systems. The emergence of open-weight models from China—which typically have minimal safeguards and can be downloaded and deployed by anyone—has intensified policy anxieties around AI security and foreign competition.