Technology and Science12 September, 20263 min read
Beijing Rejects Anthropic Claims DeepSeek, Xiaomi, Moonshot AI, Alibaba Distilled Claude Prompts
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Anthropic alleges Alibaba, DeepSeek, Moonshot AI used Claude to train their models. China rejects the claims, calling for AI for good and accusing Anthropic of distorting facts.
Beijing Rejects Anthropic Claims
Beijing rejected Anthropic’s claims that Chinese AI labs, including DeepSeek, Xiaomi, Moonshot AI and Alibaba, engaged in distillation attacks against Claude, with Chinese Foreign Ministry spokesperson Mao Ning telling a regular news conference that Beijing “advocates AI for good.” Mao Ning said Beijing “firmly opposes attempts to throw mud at China by distorting facts,” after Anthropic published its report titled Detecting and Countering Misuse of AI: September 2026 on Thursday. Anthropic’s report alleged that Chinese entities rerouted inbound user prompts to Claude to harvest chain-of-thought reasoning and train their own architectures, and that unauthorized pass-throughs exposed sensitive third-party data.
The allegations were framed as “distillation,” which the sources describe as a technique where a less powerful AI system learns from responses generated by a more advanced model. The dispute follows Anthropic’s threat intelligence report that covers misuse detected between December 2025 and August 2026, according to International Business Times.
Claude Routing Allegations
Anthropic accused Alibaba of the largest distillation operation it measured, saying it detected more than 151 million Claude exchanges between May and July, with activity peaking at nearly 3 million exchanges a day from more than 3,500 fraudulent accounts. In the same report, Anthropic said operators affiliated with Alibaba targeted reasoning transcripts from Anthropic's Opus 4.6 and 4.7 models and converted them into data for supervised fine-tuning. International Business Times also said DeepSeek was linked to more than 12.1 million exchanges over a 14-day period in July, while Moonshot AI was linked to more than 23 million Claude exchanges between May and July.
The South China Morning Post described Anthropic’s escalation as accusing Moonshot and DeepSeek of covertly routing customer requests to Claude and then showing users the responses as if they were from their own models, noting that the company said the findings “raise concerns about the misuse of user data by PRC AI labs.” The Times of India similarly reported that Moonshot was accused of secretly routing Kimi users’ requests to Claude and using some exchanges to train its models, while Beijing rejected the accusations through Mao Ning’s remarks.
Scale, Data Exposure, Fallout
Anthropic’s account, as summarized by International Business Times, said Moonshot relayed nearly 300,000 customer requests through 5,380 accounts during one 10-day period, mostly to Claude Opus models, and that some requests contained sensitive information. The report described one user likely affiliated with China’s People’s Liberation Army submitting surveillance information from hundreds of cameras in Chengdu while believing it was being processed by Kimi, and another user at a major Chinese state-owned enterprise submitting internal code and active credentials belonging to several companies. International Business Times also said Anthropic linked Xiaomi to more than 400,000 exchanges during a 20-day period in March and April, and that some traffic included names, contact information and corporate data belonging to users of third-party model-routing services.
The Times of India added that Anthropic said the operations involved more than 151 million exchanges with Claude by Alibaba alone between May and July, and that the activity peaked at nearly 3 million exchanges a day and involved more than 3,500 accounts Anthropic described as fraudulent. Across the dispute, Beijing’s response remained anchored in Mao Ning’s rejection of “distorting facts,” while the allegations themselves were presented as covering misuse detected between December 2025 and August 2026 across seven areas of AI misuse, including cyber operations, surveillance, fraud, weapons development and model distillation.