The United States and China are locked in a dispute over a technique known as AI model distillation, with the U.S. government calling it intellectual property theft and Chinese companies and officials framing it as a natural part of competition. The process involves taking insights from large language models (LLMs)—complex AI systems capable of generating text, answering questions, and performing other tasks—and using them to train smaller, more efficient models. This has become a focal point in the larger global race to develop advanced artificial intelligence. On April 23, 2026, the Trump administration accused Chinese entities of systematically copying American AI models, a claim that Beijing denied. In response, DeepSeek, a prominent Chinese AI company, released its V4 model the next day. The model was built using a method called step-by-step distillation, introduced in September 2023, which allows for smaller models to be trained with less data while still outperforming the original large models in specific tasks. Jensen Huang, CEO of Nvidia, one of the world's leading makers of AI chips, argued that distillation is not theft but rather a form of competition. He stated that companies have the right to test and learn from other products, and if a company is unhappy with this, it can simply stop providing access to its services. Huang emphasized that competition drives progress in all fields, including AI. In July, U.S. Treasury Secretary Scott Bessent labeled distillation as theft and warned that foreign companies using the technique to extract capabilities from U.S.-developed models could face sanctions. Huang again countered this view, insisting that distillation is a legitimate part of the competitive landscape. Earlier this month, the U.S. Cybersecurity and Infrastructure Security Agency (CISA) accused Chinese AI companies of conducting large-scale distillation campaigns, which it claimed violated the terms of service of American companies. A report by Anthropic, an AI research company, published between December 2025 and August 2026, alleged that Chinese labs such as Alibaba, Moonshot, and DeepSeek used millions of interactions with Anthropic’s Claude AI system to train their models through distillation. The report suggested that some of these interactions included sensitive information from individual users, corporations, and even state-linked actors, potentially violating privacy laws and the companies' own terms of use. U.S. federal authorities have accused China of systematically distilling American AI models using advanced systems. Evidence of this was found in Chinese models after complaints from Anthropic and OpenAI, who claimed Chinese companies had stolen their models using the distillation technique.