U.S. Accuses Chinese AI Firms of Systematic Theft of American Technology

In a recent alert, the FBI, National Security Agency, and Cybersecurity and Infrastructure Security Agency have detailed accusations against six leading Chinese AI companies for systematically extracting advanced AI models from American firms. This practice, known as distillation, has been ongoing since 2024 and involves companies like Anthropic, OpenAI, Google, and xAI. The U.S. government has increasingly voiced concerns that Chinese AI models are built by distilling American technology, which often violates the terms of service of the original models.

The report names six Chinese companies: DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI, and provides specific examples of how these companies have allegedly distilled various capabilities from American models. The findings suggest that the scale and sophistication of these operations indicate that distillation is not merely a supplementary method but rather a core aspect of these companies' AI development strategies.

Historically, the U.S. has accused China of intellectual property theft, and this report suggests that the distillation practices may occur with the awareness of the Chinese government, although it stops short of claiming direct involvement from intelligence agencies. The White House has framed the competition in AI development as a race against China, with President Donald Trump scheduled to meet with Chinese President Xi Jinping soon.

The report also includes recommendations for American companies to enhance user verification processes and to collaborate on identifying suspicious behaviors related to AI model usage. In previous instances, companies like OpenAI and Google have raised alarms about cloning attempts, with OpenAI specifically accusing DeepSeek of basing its model on ChatGPT.

Michael Kratsios, director of the White House Office of Science and Technology Policy, has emphasized the critical nature of this issue, stating that such distillation practices allow Chinese firms to benefit from American innovation without adhering to safety standards. He has called for a distinction between legitimate AI distillation for efficiency and large-scale operations aimed at stealing proprietary technology, which he deems unacceptable.