China accuses US of AI hegemonism, threatens retaliation over Moonshot distillation claims

China fights back in AI spat with claim US AI companies distil Chinese models

China's government has vowed to "take all necessary measures" if the US government imposes new sanctions on AI companies. That threat came on Monday at a media Q&A staged by China's Ministry of Commerce, during which local reporters asked about allegations that Chinese AI companies have distilled leading American AI models. The source of those allegations was Donald Trump's Assistant for Science and Technology, Michael Kratsios, whose stance was backed by US Treasury Secretary Scott Bessent. China always denies that its military, intelligence services, and private industry steal any information or technology, and characterizes allegations of theft as propaganda intended to discredit the nation. The Ministry's response to the Trump administration's allegations re-uses those arguments. It then suggests that Chinese AI companies didn't have enough time to distill the latest frontier models, because the US companies that developed them released them mere days before Chinese companies published their own code. Beijing also asserts "some Chinese models are already in a leading position" and that allegations of distillation are therefore unfounded. Which leaves China arguing that its AI companies are good, but not so good they can quickly distill frontier models and re-release them with Chinese characteristics. The Ministry then throws out its own allegation: "It is understood that many American AI companies have distilled Chinese models in their research and training." The allegation is plausible but not necessarily a sign that US companies are abusing intellectual property. That argument comes from the Trump administration's Kratsios, who last week argued that AI distillation can be a legitimate technique when "used to create smaller, more efficient models." American model-makers could therefore distill hundreds of Chinese open source models without behaving badly. The Ministry's post doesn't specify how China's government might respond to sanctions but ends with a call for the US to "jointly promote the positive development of artificial intelligence, and benefit both countries and the world by advancing and improving AI governance." That's a reference to the World Artificial Intelligence Cooperation Organization (WAICO), a 29-member organization that China recently convened to promote AI governance and deployment. China has also called for AI models to be open sourced, to advance innovation and in a spirit of cooperation. The China/US AI spat is taking place at the same time that stateside debate about AI regulation heated up after 77 technology companies signed an open letter calling on US lawmakers to support open weights models. China and a big section of the tech industry are therefore on the same side on at least one of the many issues AI raises. ®

China accuses US of 'AI hegemonism', threatens countermeasures over potential probes

BEIJING, July 27 (Reuters) - China's commerce ministry on Monday accused the United States of "AI hegemonism" and threatened countermeasures after senior U.S. officials said Chinese AI companies could face investigations, sanctions and trade restrictions over the alleged theft of U.S. technology. The ministry said Washington was threatening Chinese companies with punishment based on allegations they used "distillation" to copy advanced U.S. AI models, despite what it called a lack of factual or legal grounds. Model distillation is a widely used AI training technique in which developers use the outputs of a more capable model to train or improve another system. U.S. officials say they distinguish between legitimate distillation and large-scale extraction that amounts to intellectual property theft. 'ALL NECESSARY MEASURES' "For any action that causes substantive harm to ⁠Chinese interests, China will take all necessary measures to firmly safeguard its legitimate rights and interests," a ministry spokesperson said in a statement. The comments mark an escalation in a dispute centred on Beijing-based Moonshot AI, whose recently released Kimi K3 model has drawn attention for its coding capabilities and intensified debate in Washington over whether Chinese developers are copying U.S. models or rapidly closing the technological gap through their own research. The controversy echoes the reaction in the U.S. to the rise of Chinese startup DeepSeek last year. White House Office of Science and Technology Policy Director Michael Kratsios said last week the U.S. government had information indicating Moonshot distilled Anthropic's Claude Fable 5 model to develop Kimi K3. Kratsios alleged Moonshot had built a sophisticated internal platform to conduct large-scale distillation of U.S. models and switched between multiple methods ⁠of access to avoid detection. He also alleged the company had acquired servers equipped with Nvidia's GB300 chips and accessed such systems in Thailand, likely to train its models. U.S. Treasury Secretary Scott Bessent later warned that Chinese companies could face financial sanctions or placement on the Commerce Department's Entity List, which restricts access to U.S. technology. "We support open-source AI and the innovation it unlocks. But open source is not open season on American IP," ⁠Bessent wrote on social media platform X. "When PRC (Chinese) firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table," he said. Moonshot has rejected suggestions that Kimi K3's performance was achieved via distillation, telling China's National Business Daily ⁠last week that its performance gains came from original changes to underlying architecture. Placement on the Entity List could severely restrict Moonshot's access to U.S. semiconductors, software and cloud services. The U.S. used the trade blacklist as a central part of its campaign against Chinese telecommunications ⁠equipment maker Huawei beginning in 2019. Anthropic said in February it had identified more than 3.4 million interactions with its Claude models linked to Moonshot. It said hundreds of fraudulent accounts were used to target capabilities including reasoning, coding, data analysis, tool use, computer vision and the operation of computers. Reporting by Eduardo Baptista and Beijing Newsroom. Editing by Alexandra Hudson and Mark Potter Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Disrupted Eduardo Baptista Thomson Reuters Eduardo Baptista is a Senior Correspondent for Reuters based in Beijing, covering China's technology, space, and automotive industries. He has led enterprise and investigative reporting on China's military-linked companies, artificial intelligence and semiconductor supply chains, as well as macroeconomic and industrial policy. Baptista has reported from China for nearly a decade and holds a BA in History from the University of Cambridge.

China accuses the US of 'AI hegemonism' and threatens countermeasures

Beijing's commerce ministry warned of "all necessary measures" after Washington raised the prospect of sanctions over alleged copying of Anthropic's Claude. China's commerce ministry accused the United States of "AI hegemonism" on Monday and warned that it would retaliate, an escalation in a quarrel that started with an accusation of copying and has since pulled in chips, sanctions, and a pair of senior White House officials. The ministry said Washington was threatening Chinese companies with punishment on the basis of allegations alone, and it cast the whole affair as an effort to hold back China's technology sector rather than a genuine grievance about intellectual property. The trigger was distillation, the practice of training a weaker model on the outputs of a stronger one. White House Office of Science and Technology Policy director Michael Kratsios alleged that the Beijing firm Moonshot AI had distilled Anthropic's Claude to build its Kimi K3 system, and that it had done so at industrial scale through a purpose-built internal platform designed to dodge detection. The White House had already put China on notice over the practice, though enforcing any of it remains the hard part. Treasury Secretary Scott Bessent went further, warning that firms caught running covert distillation attacks could face financial sanctions or a place on the Commerce Department's Entity List, which would cut them off from American chips and software. That sanction threat is what appears to have prompted Beijing's response. The ministry's language was measured but firm. "For any action that causes substantive harm to Chinese interests, China will take all necessary measures to firmly safeguard its legitimate rights and interests," an unnamed spokesperson said, without spelling out what those measures might be. The ministry also rejected the underlying charge, arguing that distillation is a widely used technique, and that the US had produced neither factual nor legal grounds for the punishment it was threatening. The complaint that started all this came from Anthropic itself. Earlier this year the company told US lawmakers that three Chinese firms, DeepSeek, Moonshot, and MiniMax, had generated more than 16 million interactions with Claude using roughly 24,000 fraudulent accounts, extracting capabilities in reasoning, coding, and computer vision that they then folded into their own systems. It later named Alibaba's Qwen lab as the source of an even larger distillation campaign, and it has used the findings to argue for tighter chip export controls, on the logic that restricting access to advanced semiconductors constrains both direct training and the distillation that feeds off it. Moonshot has denied the charge. The company said the gains in Kimi K3 came from its own architectural work rather than from copying a rival, a line that echoes the broader Chinese position that Washington is dressing up commercial protectionism as a security concern. Whether US regulators accept that is a separate question, and one the Entity List does not require them to answer before acting. The chip dimension runs underneath all of it. Washington has spent the past year closing loopholes that let Nvidia's most advanced processors reach Chinese buyers through overseas subsidiaries, and reporting tied to the Moonshot case describes the company sourcing Nvidia GB300 chips and routing work through servers in Thailand. Neither Beijing nor Washington has confirmed those specifics on the record, and Anthropic's own tally of the traffic it attributes to Moonshot has not been independently audited. Notably absent from Monday's exchange was any mention of a Section 301 trade investigation, the tariff mechanism the US has reached for against foreign technology rules elsewhere. This fight is narrower, and for now it is aimed squarely at a handful of Chinese AI labs and the models they have shipped. What comes next is procedural. Bessent's warning describes a possible sanction, not one that has been imposed, and no company has yet been added to the Entity List over distillation. The ministry's countermeasures remain unspecified. For the moment both sides have staked out positions and left themselves room, which is its own kind of signal.

China accuses US of 'AI hegemonism', threatens countermeasures over potential probes

The ministry said Washington was threatening Chinese companies with punishment based on allegations they used "distillation" to copy advanced US AI models, despite what it called a lack of factual or legal grounds. China's commerce ministry on Monday accused the United States of "AI hegemonism" and threatened countermeasures after senior US officials said Chinese AI companies could face investigations, sanctions and trade restrictions over the alleged theft of US technology. The ministry said Washington was threatening Chinese companies with punishment based on allegations they used "distillation" to copy advanced US AI models, despite what it called a lack of factual or legal grounds. Model distillation is a widely used AI training technique in which developers use the outputs of a more ⁠capable model ⁠to train or improve another system. US officials say they distinguish between legitimate distillation and large-scale extraction that amounts to intellectual property theft. 'All necessary measures' "For any action that causes substantive harm to Chinese interests, China will take all necessary measures to firmly safeguard its legitimate rights and interests," a ministry spokesperson said in a statement. The comments mark an escalation in a dispute centred on Beijing-based Moonshot AI, whose recently released Kimi K3 model has drawn attention for its coding capabilities and intensified debate in Washington over whether Chinese developers are copying US models or rapidly closing ⁠the technological gap through their own research. The controversy echoes the reaction in the US to the rise of Chinese startup DeepSeek last year. White House Office of Science and Technology Policy Director Michael Kratsios said ⁠last week the US government had information indicating Moonshot distilled Anthropic's Claude Fable 5 model to develop Kimi K3. Kratsios alleged Moonshot had built a sophisticated internal platform to conduct large-scale distillation of US models and switched between multiple methods of access to avoid detection. He also alleged the company had acquired servers equipped with Nvidia's GB300 chips and accessed such systems in Thailand, likely to train its models. US Treasury Secretary Scott Bessent later warned that Chinese companies could face financial sanctions or placement on the Commerce Department's Entity List, which restricts access to US technology. "We support open-source AI and the innovation it unlocks. But open source is not open season on American IP," Bessent wrote on social media platform X. "When PRC (Chinese) firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be ⁠on the table," he said. Moonshot has rejected suggestions that Kimi K3's performance was achieved via distillation, telling China's National Business Daily last week that its performance gains came from original changes to underlying architecture. Placement on the Entity List could severely restrict Moonshot's access to US semiconductors, software and cloud services. The US used the trade blacklist as a central part of its campaign against Chinese telecommunications equipment maker Huawei beginning in 2019. Anthropic said in February it had identified more than 3.4 million interactions with its Claude models linked to Moonshot. It said hundreds of fraudulent accounts were used to target capabilities including reasoning, coding, data analysis, tool use, computer vision and the operation of computers.

China accuses U.S. of 'AI hegemonism,' threatens countermeasures over potential probes

BEIJING - China's commerce ministry on Monday accused the United States of "AI hegemonism" and threatened countermeasures after senior U.S. officials said Chinese AI companies could face investigations, sanctions and trade restrictions over the alleged theft of U.S. technology. The ministry said Washington was threatening Chinese companies with punishment based on allegations they used "distillation" to copy advanced U.S. AI models, despite what it called a lack of factual or legal grounds. Model distillation is a widely used AI training technique in which developers use the outputs of a more capable model to train or improve another system. U.S. officials say they distinguish between legitimate distillation and large-scale extraction that amounts to intellectual property theft.

China Accuses US of AI Hegemony, Dismisses White House Threats of Sanctions

The Chinese appear confident that on this occasion, the US threats are hollow as their recent open-weight AI models were well received by Corporate America China has accused the United States of "AI hegemonism" following reports of Donald Trump aides seeking a ban on Chinese open-weight AI models. Especially after American enterprises appeared to have embraced in recent times due to its lower costs. Beijing has threated counter-measures in case Washington began investigations or placed sanctions or trade restrictions. At the root of the latest fracas is the latest open-weight AI model launched by Moonshot, which received a beyond-warm welcome from industry experts as well as customers in the US. These supporters also included top names like Satya Nadella of Microsoft who has been articulating his views against closed models in recent times. Nadella has said that closed AI models were like Trojan Horses where customers paid twice - once for use of the product and again by way of data they shared involuntarily with the AI model in the form of prompts and other stuff during their interaction with the chatbots. The gripe among US models around their Chinese counterparts is that the latter was created using distillation as a means of using existing AI models from the United States. Beijing now notes that American accusations appear to be based on hearsay from model-makers like Anthropic and lacked factual or legal grounds. Of course, China also notes that model distillation has been a widely used AI training technique among developers where they use the outputs of an earlier model to train and enhance the capabilities of a new one. However, US authorities have countered this claim suggesting that they can distinguish between legitimate distillation and largescale extraction of intellectual property. "For any action that causes substantive harm to Chinese interests, China will take all necessary measures to firmly safeguard its legitimate rights and interests," a ministry spokesperson said in a statement. The escalation features a dispute around Moonshot AI's Kimi K3 model that drew attention over its coding capabilities being almost close to existing models of US AI companies. Last week, we had reported about how Kimi K3 became Trump's biggest dilemma beyond the Iran war and how the answer was provided for by the American tech industry. Be that as it may, from the Chinese point of view, President Trump and his cronies' reaction to a Chinese AI model is just indicative of one thing - that the White House sees Beijing as a threat to its hegemonistic plans of using AI for their larger MAGA plan.

China accuses US of 'AI hegemonism', threatens countermeasures over potential probes

BEIJING, July 27 (Reuters) - China's commerce ministry on Monday accused the United States of "AI hegemonism" and threatened countermeasures after senior U.S. officials said Chinese AI companies could face investigations, sanctions and trade restrictions over the alleged theft of U.S. technology. The ministry said Washington was threatening Chinese companies with punishment based on allegations they used "distillation" to copy advanced U.S. AI models, despite what it called a lack of factual or legal grounds. Model distillation is a widely used AI training technique in which developers use the outputs of a more capable model to train or improve another system. U.S. officials say they distinguish between legitimate distillation and large-scale extraction that amounts to intellectual property theft. 'ALL NECESSARY MEASURES' "For any action that causes substantive harm to Chinese interests, China will take all necessary measures to firmly safeguard its legitimate rights and interests," a ministry spokesperson said in a statement. The comments mark an escalation in a dispute centred on Beijing-based Moonshot AI, whose recently released Kimi K3 model has drawn attention for its coding capabilities and intensified debate in Washington over whether Chinese developers are copying U.S. models or rapidly closing the technological gap through their own research. The controversy echoes the reaction in the U.S. to the rise of Chinese startup DeepSeek last year. White House Office of Science and Technology Policy Director Michael Kratsios said last week the U.S. government had information indicating Moonshot distilled Anthropic's Claude Fable 5 model to develop Kimi K3. Kratsios alleged Moonshot had built a sophisticated internal platform to conduct large-scale distillation of U.S. models and switched between multiple methods of access to avoid detection. He also alleged the company had acquired servers equipped with Nvidia's GB300 chips and accessed such systems in Thailand, likely to train its models. U.S. Treasury Secretary Scott Bessent later warned that Chinese companies could face financial sanctions or placement on the Commerce Department's Entity List, which restricts access to U.S. technology. "We support open-source AI and the innovation it unlocks. But open source is not open season on American IP," Bessent wrote on social media platform X. "When PRC (Chinese) firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table," he said. Moonshot has rejected suggestions that Kimi K3's performance was achieved via distillation, telling China's National Business Daily last week that its performance gains came from original changes to underlying architecture. Placement on the Entity List could severely restrict Moonshot's access to U.S. semiconductors, software and cloud services. The U.S. used the trade blacklist as a central part of its campaign against Chinese telecommunications equipment maker Huawei beginning in 2019. Anthropic said in February it had identified more than 3.4 million interactions with its Claude models linked to Moonshot. It said hundreds of fraudulent accounts were used to target capabilities including reasoning, coding, data analysis, tool use, computer vision and the operation of computers. (Reporting by Eduardo Baptista and Beijing Newsroom. Editing by Alexandra Hudson and Mark Potter)