미 “중국 6개 AI기업, 미국 기술 ‘악의적 증류’”…정상회담 앞 긴장 고조

The United States government has publicly denounced major Chinese artificial intelligence (AI) companies for "aggressively and maliciously refining" advanced American models to secure technology, according to a joint cybersecurity advisory issued by the National Security Agency (NSA), Federal Bureau of Investigation (FBI), and Cybersecurity and Infrastructure Security Agency (CISA).

This warning comes in anticipation of the upcoming summit between U.S. President Donald Trump and Chinese State Councilor and Vice President Xi Jinping. Tensions between the two nations surrounding artificial intelligence (AI) supremacy are escalating as the summit approaches.

On August 8, the three U.S. agencies released a cybersecurity advisory claiming that "Chinese-based AI companies have launched industrial-scale, aggressive, malicious, and targeted refining activities" to extract limited exclusive functions and capabilities from American AI models. The agencies warn that such malicious actions and techniques must be disclosed and their potential impact mitigated.

The term "refining" refers to the technique of using high-performance large AI models' outputs to train smaller or new AI models, which is commonly employed in AI research and development. U.S. authorities argue that Chinese companies, instead of adhering to standard refining, circumvent regional restrictions and bypass American companies' usage agreements and safety measures to conduct industrial-scale refining.

The NSA, FBI, and CISA point to six organizations – DeepScribe, Medical Speech AI, Alibaba, Minimax, StepFun, and Zephi – as using refining to extract billions of tokens from open-source AI models such as OpenAI's GPT, Anthropic's Claude, and Google's Gemini.

The U.S. government believes that Chinese companies' aggressive refining activities will strengthen China's military capabilities as they can develop advanced AI models without incurring the substantial research and development costs, computing resources, and power required by the U.S. and its allies. The NSA highlights that the outcome of this refining will not only enhance China's military and cyber attack capabilities against the U.S. and its allies but also intensify the competition for advanced AI between the two nations.

As the United States ties China's AI development methods to national security threats, the issue is expected to be a key topic of discussion at the upcoming U.S.-China summit.