The White House has publicly claimed that Chinese artificial intelligence (AI) company Moonshot AI unlawfully used key technologies from American AI company Anthropic on a large scale.

The U.S. government believes the Chinese company replicated the performance of American AI models through a 'distillation' technique and utilized Nvidia's advanced AI servers, with expectations that U.S.-China AI technology conflicts will intensify further.

▲ White House: "Moonshot AI Engaged in Large-Scale Theft of Anthropic AI Technology"

According to BBC on the 23rd (local time), Michael Kratsios, White House advisor for science and technology policy, revealed through social media X that the U.S. government has identified signs of large-scale AI technology theft by Moonshot AI.

He claimed that Moonshot AI extracted the performance of Anthropic's AI model 'Fable AI' using distillation techniques and applied it to developing its latest model 'Kimi K3'.

Distillation is a technology that learns answers generated by high-performing AI models repeatedly to quickly enhance the performance of relatively smaller AI models.

While the industry also uses it as a common AI development technique, using it without permission from the original developer can raise intellectual property infringement concerns.

▲ "Also Utilizing Nvidia's Most Advanced Servers"

Advisor Kratsios claimed that Moonshot AI appears to have accessed AI servers based on Nvidia's latest GB300 Grace Blackwell platform.

Since 2022, the U.S. has restricted Nvidia's cutting-edge AI chip exports to China over concerns about strengthening Chinese military capabilities.

Since then, the U.S. and major allied nations have continuously strengthened enforcement to block the smuggling of advanced AI chips into China.

▲ Suspicions Raised Following Kimi K3 Launch

Kimi K3, revealed by Moonshot AI last week, received evaluations that Chinese AI technology has significantly narrowed the gap with the United States, drawing attention from the global AI industry.

Moonshot AI has also promoted that Kimi K3 can compete with the highest-level AI models in the United States.

However, the U.S. government places weight on the possibility that such performance improvements are the result of unlawfully using the performance of American AI models rather than independent research and development.

▲ U.S.: "Will Strictly Enforce Against Industry-Scale Technology Theft"

These suspicions were revealed shortly after remarks by U.S. Treasury Secretary Scott Bessent.

On the 22nd, Secretary Bessent announced that the U.S. government would investigate whether Chinese AI companies have stolen technology, and on the 23rd, he warned that companies infringing intellectual property rights of American companies through industry-scale distillation attacks would face sanctions under consideration.

He emphasized that "the United States supports open-source AI, but open-source does not mean a license for unauthorized use of American intellectual property."

▲ Anthropic Also Calls for Lawsuit Against Alibaba

Anthropic also claimed in June that Chinese Alibaba unlawfully utilized its Claude AI model through distillation techniques.

At that time, Anthropic urged the U.S. Congress to strengthen sanctions against related companies and establish legal measures to prevent the outflow of American AI technology.

The White House also announced in April its plan to expand cooperation with American AI companies to prevent industrial-scale technology theft by overseas competitors.

▲ China: "Opposes Politicization of Technology Issues"

The Chinese government has not issued an official position on these suspicions.

However, the Chinese Embassy to the U.S. cited existing remarks by Lin Jian, spokesperson of China's Foreign Ministry, to express opposition to politicizing technology and trade issues.

Spokesperson Lin emphasized that China's AI industry has grown through independent research and development and international cooperation, and AI advancement should aim for common contribution and mutual benefit.

This controversy is notable for being raised ahead of an expected summit between U.S. President Donald Trump and Chinese President Xi Jinping in September.

In particular, Moonshot AI is scheduled to release Kimi K3 as open-source on the 27th, allowing developers worldwide to freely utilize it.