China's star AI labs routed user requests to Claude at least 35 million times in the summer: Anthropic
Anthropic said Alibaba, Moonshot, Zhipu, DeepSeek, and Xiaomi all tried to use Claude to train their models in various ways.
TL;DR
- Anthropic reported that top Chinese AI labs launched nearly 190 million distillation attacks against Claude from May to July.
- Distillation attacks involve using outputs from advanced AI models to train less advanced ones.
- Alibaba, Moonshot AI, DeepSeek, Zhipu, and Xiaomi were accused of rerouting user requests to Claude.
- Some rerouted requests exposed sensitive data from Chinese and Russian governments and militaries.
- Alibaba reportedly conducted the largest attack, with over 151 million exchanges from fraudulent accounts.
- Anthropic has added safeguards, reduced detailed reasoning transcripts, and will require identity verification for users from China, Russia, or Iran.
- This follows a previous accusation by Anthropic against Alibaba in June for similar attacks.