NVIDIA's Huang: AI Model Distillation Is Competition, Not Theft

Huo Xing Finance reports that NVIDIA CEO Jensen Huang, in an interview with CNBC’s Squawk Box on Monday, stated that using competitors’ products for training or learning constitutes competition, and he rejected the characterization of AI model distillation as theft. When asked whether distillation is not theft, Huang replied that it is simply competition. He noted that people can test others’ products, and NVIDIA’s own products are sometimes disassembled down to their bare bones to understand how they work; he would prefer others not learn from NVIDIA’s products, but competition ultimately makes everything better—if one does not wish others to use their products, they can identify customers and disable services. U.S. officials have accused Chinese AI companies of using distillation—training models on the outputs of other models—to extract capabilities from U.S. systems. In July, U.S. Treasury Secretary Scott Bessent labeled this practice as theft and threatened sanctions against overseas companies that use distillation to extract capabilities from U.S. models. The White House did not immediately respond to CNBC’s request for comment, and U.S. officials are considering measures against overseas companies using distillation. Earlier this month, the U.S. Cybersecurity and Infrastructure Security Agency accused Chinese AI companies of engaging in industrial-scale knowledge distillation activities in violation of U.S. companies’ terms of use. AI company Anthropic also reported earlier this month that it had detected Alibaba’s Qwen series models and DeepSeek conducting illegal distillation. China has denied these allegations.