Claude AI: Anthropic's "Spyware" Sparks China Tensions

Complaints have surfaced on Reddit about a mechanism embedded within Claude Code, appearing to covertly flag users based in China or with ties to Chinese AI labs. This mechanism reportedly detects signs like timezone and the use of proxy URLs. One Redditor expressed concern that Anthropic could use such spyware for more malicious ends. Anthropic responded on X, with a technical staff member stating it was an experiment launched in March to prevent account abuse and protect against distillation. They added that the company had intended to remove it and that it should be fully rolled back with the redeployment of Fable 5. Anthropic has previously accused Chinese AI labs, including DeepSeek, Moonshot, MiniMax, and Alibaba, of illicitly using Claude to train new AI models. This practice, known as distillation, has become a sensitive point in the AI competition between the U.S. and China. The White House aims to address campaigns that extract capabilities from American AI models. This situation highlights the ongoing tensions in the global AI landscape.