The Chinese online trading company Alibaba is now banning its employees from using Anthropic’s AI model Claude. Employees are also instructed to delete all Claude models from their work computers. The background is said to be a hidden function in the Claude source code that allows Anthropic to see whether a user is accessing from China or is connected to a Chinese AI developer, the employees report.
Read more after the ad
They spoke to the IT news portal The Information. Although Anthropic generally prohibits the use of its models in China, Chinese companies find ways to still use the models undetected. Alibaba is thus prohibiting its workforce from doing something that Anthropic doesn’t allow with its AI anyway.
Serious allegations against Alibaba
Last month, Anthropic accused Alibaba of AI model theft, so to speak. Alibaba is said to have attempted to illegally access Claude’s logic through mass access. Anthropic is therefore calling on the US Congress to take measures to prevent such attacks, as can be seen from a letter to American senators published by the Financial Times. Alibaba would therefore have resorted to the method of AI knowledge distillation, in which developers of a new AI have it trained with knowledge from an already existing model.
In June, the US government cited this scenario, among others, as a reason to ban Anthropic’s Fable and Mythos models for non-US citizens. Alibaba initially did not comment on the allegations. However, the current reports from its employees are a clear indication that Alibaba is illegally using Anthropic’s AI. Anthropic had previously accused Chinese AI manufacturers DeepSeek, Moonshot AI and MiniMax of distillation.
Alibaba also recently filed a lawsuit against the Pentagon. The Chinese company had put this on a list of suspected military companies. The lawsuit is now intended to reverse this.
Read also
(nen)
