China's Moonshot AI Accused of Anthropic IP Theft Amid US Scrutiny

Top US technology officials have accused China-based artificial intelligence (AI) startup Moonshot AI of intellectual property theft from its American rival, Anthropic. Michael Kratsios, Director of the White House Office of Science and Technology Policy, stated that the US possesses information indicating Moonshot AI illicitly 'distilled' Anthropic's Fable model to develop its new K3 model. These allegations further inflame the already tense US-China technology rivalry, raising serious questions about the legitimacy of Beijing's rapid advancements in the AI sector.

According to Michael Kratsios' statements, Moonshot AI allegedly developed a sophisticated internal platform designed to mimic the inner workings and outputs of Anthropic's Fable model. This platform is reportedly engineered to conduct large-scale distillation attacks against US models and rapidly switch access methods to avoid detection. Furthermore, Moonshot AI is accused of acquiring NVIDIA's advanced GB300 chips via Thailand, circumventing US export controls, and using these chips to train its AI models. Sarah Heck, Head of Public Policy at Anthropic, characterized these actions as “IP theft and industrial espionage,” emphasizing the serious national security risks they pose to the United States.

Following these accusations, the Trump administration issued a warning regarding potential sanctions against Chinese AI companies. US Treasury Secretary Scott Bessent stated on Fox Business that if overseas models are found to be stealing intellectual property from American companies, the US has the authority to impose sanctions and export restrictions on those entities. Bessent specifically highlighted a focus on “covert, industrial-scale distillation attacks” aimed at replicating US AI models. The Bureau of Industry and Security, operating within the US Department of Commerce, is also reportedly investigating Chinese AI companies, including Moonshot AI, for their access to advanced US AI chips.

These developments have sent significant ripples through global technology markets, underscoring the critical importance of competition and intellectual property protection in the artificial intelligence sector. This situation could indirectly impact chip manufacturers like NVIDIA (NVDA) and heightens the significance of restrictions on access to high-performance AI chips. The US threat of sanctions against Chinese AI firms has the potential to escalate the technology war between the two nations beyond semiconductor supply chains, into more complex domains such as AI model training methodologies and intellectual property rights.

In a broader economic and political context, US lawmakers are intensifying their scrutiny of American companies' increasing adoption of Chinese-developed AI models, citing national security and cybersecurity concerns. The House Committee on Homeland Security and the House Select Committee on China have launched a joint investigation into the embrace of China-developed AI technologies. In response, Liu Chang of the Chinese embassy in Washington dismissed Kratsios' comments as “entirely unfounded,” asserting that China respects intellectual property protections.

Analysts and market observers suggest that such allegations will fuel further debate within the AI industry regarding the distinction between “legitimate distillation” and “illicit intellectual property theft.” Some experts argue that distillation is a widely used technique in AI development and that Chinese firms possess strong research teams. However, US officials maintain that Moonshot AI's actions constitute large-scale, covert industrial distillation aimed at stealing proprietary US technology and undermining American research. In the coming period, concrete actions by the US in response to these allegations and potential retaliatory measures from China will be key factors shaping the trajectory of global AI competition.