U.S. Administration Divided Over Measures Against Chinese AI Distillation Attacks

The Trump administration is experiencing significant internal discord over how to address the rapid rise of Chinese AI models. According to a report by Hugo Lowell of WIRED, this conflict has escalated between the White House and the Department of Commerce. While the White House advocates for stringent regulations on Chinese AI, the Commerce Department argues that such measures are impractical.

The Source of Division

This turmoil was directly triggered by the release of China’s Moonshot AI model, Kimi K3, last week. The model demonstrates capabilities comparable to cutting-edge models developed by U.S. firms such as Anthropic and OpenAI, sending shockwaves through the U.S. government. Additionally, last month, Anthropic accused Chinese tech giant Alibaba of executing the “largest known distillation attack” to date.

In response to these incidents, senior officials at the White House are deliberating measures to counter the method known as distillation. Distillation involves using the outputs of high-performance U.S. AI models as training data, allowing Chinese companies to efficiently train their own models.

The Commerce Department’s Pragmatic Approach

While the White House leans toward regulatory tightening, the Department of Commerce is exploring alternative strategies. As the overseer of the Bureau of Industry and Security (BIS), which manages export controls, the Commerce Department’s influence is growing. Commerce Secretary Howard Rutnick is reportedly considering ways to incentivize leading U.S. research institutions to develop open-weight models, aiming to offset China’s influence.

According to the report, Secretary Rutnick has met with leaders of several AI research institutions to seek a middle ground solution that does not solely rely on stricter regulations. Skeptical of the effectiveness of such regulations, he believes that enhancing U.S. technological prowess would be a more realistic approach to maintaining competitiveness.

Alternatives to an Executive Order

U.S. Treasury Secretary Scott Besant referred to distillation, often conducted covertly, as “intellectual property theft” during an interview with Fox Business, hinting at potential sanctions. In fact, the Treasury Department recently issued warnings of sanctions against Chinese AI companies in connection with Moonshot AI’s suspected distillation practices.

While some form of presidential action is anticipated, insiders suggest that it is unlikely to take the form of a conventional executive order. Instead, alternative measures are under consideration.

Risks of Open-Weight Models

One of the administration’s greatest policy challenges is the rise of “open-weight models” being developed in China, in addition to traditional closed models. These models pose heightened national security risks because they lack safeguards to prevent their misuse in infiltrating government systems or critical infrastructure, and they are freely downloadable by anyone.

Anthropic’s models have already garnered attention for their ability to identify vulnerabilities in government systems and critical infrastructure. The controversies surrounding these models have evolved into a broader policy debate over distillation-related risks.

Editorial Opinion

In the short term, the U.S.-China standoff over AI model distillation is likely to intensify. Anthropic’s accusations against Alibaba could serve as a catalyst for the announcement of some form of presidential action within the year. Measures may not be limited to Moonshot AI or Alibaba but could include strengthened restrictions on access to specific semiconductors and cloud services.

From a long-term perspective, the U.S. is expected to pursue a dual approach involving “export controls on advanced semiconductors” and “prevention of distillation attacks.” This could drive China’s AI development toward a more closed and self-contained trajectory. However, this geopolitical divide risks significantly hampering the global standardization of AI safety and international research collaboration.

The editorial board believes that the deep divide between the ideal of democratizing open-weight models and the realities of national security will become the greatest challenge for future AI governance. The critical question remains whether regulations will be effective or simply push development underground. The ability of policy to keep pace with technological advancements is now under scrutiny.

References

• “The White House Is Trying to Figure Out What to Do About Chinese AI”, by Hugo Lowell — Wired, 2026-07-22T21:00:00.000Z (ARR)
• Source URL: https://www.wired.com/story/the-white-house-is-trying-to-figure-out-what-to-do-about-chinese-ai/