China’s AI Heist: How to Counter Beijing’s Unauthorized “Distillation”

Foreign Affairs

The U.S.-China competition in artificial intelligence has expanded to open-weight, local AI models, where Chinese firms are gaining a significant advantage through unauthorized "distillation." Chinese companies systematically extract capabilities from frontier American AI systems by training smaller, more efficient models to mimic sophisticated U.S. ones at an industrial scale, a practice U.S.

firms cannot legally replicate. This creates a profound asymmetry, allowing Chinese models to disproportionately underpin local AI deployments globally, including in the United States, distorting the market and risking geopolitical dependence on Chinese technology. Moreover, this process bypasses critical AI safeguards, leading to models with security vulnerabilities, as demonstrated by DeepSeek-R1 and OpenClaw. Washington must respond by tightening export controls on chips enabling distillation and extending the foreign direct product rule to Chinese models incorporating extracted U.S. capabilities, while accelerating U.S. open-weight ecosystem development.

To read in details click here