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.
Anthropic revealed that several Chinese AI labs, including Alibaba, Moonshot AI, Zhipu, DeepSeek, and Xiaomi, subjected its Claude AI to millions of distillation attacks throughout the summer. These attacks aimed to train their models using Claude's outputs, with the labs together conducting nearly 190 million such assaults between May and July.
DeepSeek and Moonshot AI specifically rerouted user requests from their own models to Claude, logging 23 million and 12.1 million reroutes respectively. Some of these requests potentially exposed sensitive data, such as CCTV footage and military information. Alibaba led the pack, with over 151 million exchanges carried out by fraudulent accounts.
Anthropic responded by enhancing its safeguards against these attacks, limiting Claude's reasoning transcripts for training purposes, and implementing identity verification for users from China, Russia, or Iran, where Claude is unavailable.
Written by urgent.news from Business Insider's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
