China rejects Anthropic's accusations against AI labs
China in Beijing rejected accusations by Anthropic concerning Chinese artificial intelligence laboratories which, according to the company, may have used responses from the Claude model to train their own systems. Chinese Foreign Ministry spokesperson Mao Ning said that Beijing supports the use of artificial intelligence for good.
Beijing's response
At a regular briefing, Mao Ning said that China firmly opposes attempts to “smear China by distorting facts.” According to Anadolu Agency, the statement came after Anthropic published the report Detecting and Countering Misuse of AI: September 2026.
More current news is available on the UA.News Telegram channel Telegram.
Anthropic's claims
In its report, Anthropic claims that Chinese laboratories, including Moonshot AI and DeepSeek, carried out large-scale model distillation attacks. Distillation is a common method in which a less powerful artificial intelligence system learns from the responses of a more advanced model.
According to Anthropic, these entities redirected users' input prompts to Claude in order to obtain the model's chains of reasoning and train their own architectures. The company also mentioned DeepSeek, Xiaomi, Moonshot AI, and Alibaba among the Chinese laboratories against which it raised claims. Anthropic said that such unauthorized redirections could have inadvertently exposed confidential third-party data, including internal recordings from surveillance cameras in Chengdu.
