Beijing Rejects U.S. Allegations of AI Copying by Chinese Firms
Beijing has rejected U.S. allegations that Chinese artificial intelligence firms are copying American technology through ‘distillation,’ labeling the claims an intentional attack designed to halt China’s industrial progress. Chinese Embassy spokesperson Mr. Liu Pengyu stated that Washington is hyping the issue to curb technological advancement and urged the United States to abandon its biases. He maintained that China’s AI sector is growing entirely through its own capabilities and scientific research, demanding that technology and trade stop being used as political tools.
The friction follows accusations from the U.S. government on Tuesday targeting six Chinese AI firms, including DeepSeek, Moonshot AI, and Alibaba, for allegedly replicating American capabilities. U.S. officials claimed these companies used outputs from major U.S. models developed by Anthropic, OpenAI, Google, and SpaceX to rapidly and cheaply train their own systems. According to Washington, this industrial-scale distillation reduces research costs while potentially enhancing China’s military and cyber capabilities with government awareness.
Distillation is a recognized machine learning practice where a smaller, cost-effective model is trained using the outputs of a larger, more advanced model. While common across the global tech industry for optimizing efficiency, U.S. authorities argue that applying it to proprietary foreign frontier models crosses a line. Beijing, however, has dismissed these security concerns and accusations as entirely groundless, repeating its stance that Washington is seeking to suppress fair competition.
This escalation occurs against a backdrop of persistent technology and trade tensions between the two global superpowers. Diplomatic preparations are currently underway for bilateral talks on AI security scheduled for mid-month, followed by a high-stakes meeting between U.S. President Donald Trump and Chinese President Xi Jinping in late September.
Source: TASS
