China Rejects US Claims of ‘Industrial-Scale’ AI Model Theft

Summary

China has rejected accusations from the FBI, NSA and CISA that Chinese artificial intelligence companies have engaged in “industrial-scale” theft of US AI models. A joint US advisory said companies including DeepSeek, Alibaba and Moonshot AI had distilled billions of tokens from American models since at least late 2024, likely with awareness from the Chinese government. Model distillation trains a smaller or less capable model using outputs from a more advanced one, a practice China’s Commerce Ministry described as common throughout the industry, including among US companies. US officials argued that the scale and sophistication of the campaigns make distillation the central basis of the Chinese companies’ model development rather than a supplementary technique. China’s Commerce Ministry accused Washington of seeking a monopoly in AI and warned that suppressing Chinese companies under the pretext of targeting distillation would prompt “resolute countermeasures.” The Foreign Ministry called the US allegations unfounded and attributed China’s AI progress to technological self-reliance, while urging international cooperation. The dispute comes weeks before President Xi Jinping is expected to visit the US for a summit hosted by President Donald Trump, potentially with Chinese technology executives. The companies named in the advisory had not immediately responded to requests for comment. The article also notes that China is investing nearly $300bn in AI and robotics, while US Treasury Secretary Scott Bessent said China would not overtake the US in AI because it relies on distilling American models.