Anthropic Announces in a Report that Chinese Artificial Intelligence Companies Collect User Data Without Permission and Use Distillation Methods via Claude

Anthropic explained in its report how Chinese artificial intelligence companies collect and distill user data without permission via Claude.

As competition intensifies in the world of artificial intelligence, Anthropic, one of the leading names in the industry, presented striking details about the methods of training Chinese technology giants' artificial intelligence models in a comprehensive report published today. According to the report, it was determined that China-based companies, including important names such as DeepSeek, Xiaomi and Moonshot AI, used the questions and data asked by users to advanced models such as Claude by "distillation" method to train their own systems.

It is stated that in this process, users' personal data and sensitive company information are unknowingly transferred to the training pool of third-party models. Anthropic argues that this violates privacy laws and conflicts with the companies' own terms of use.

Chinese Companies Use Claude Data in Model Training

One of the most striking parts of the report reveals the extent of "distillation" activities. Anthropic states that Chinese developers took the questions asked by users and sent them to Claude, and turned Claude's answers into the data set of their own models. It is stated that during this data flow, names, e-mail addresses and confidential information within the company are processed in an unprotected manner. This situation poses a serious threat to data security, especially for corporate users.

Alibaba Processes Record Levels of Data

Alibaba alone is leading the largest distillation attack ever seen in the AI world, with more than 151 million transactions, according to Anthropic's analysis. Processing such a large volume of data demonstrates competitors' efforts to transfer the reasoning capabilities of advanced models like Claude to their native models. The company emphasizes that these activities are not only a technical competition, but also a process where ethical boundaries are pushed.

Moonshot AI Directs Customer Inquiries to Claude

It was determined that Moonshot AI, a company known for some of its model families, directed requests from its customers to Claude instead of solving them directly on its own infrastructure. Although this strategy is a method used to make systems appear more intelligent, moving users' data to another platform without their consent invalidates privacy agreements. Anthropic states that the data of hundreds of users is indirectly shared through this method and that this is against global data privacy standards.