Anthropic secretly embedded a tracker in Claude Code to monitor Chinese users. A security researcher discovered the hidden code, calling it spyware and a breach of trust. The tracker, using prompt steganography, collected user timezone, proxy information, and potential ties to Chinese AI labs. Anthropic engineer Thariq Shihipar confirmed it was an experiment to prevent account abuse and protect against distillation attacks. Unauthorized resellers were selling access to models at significantly reduced prices. Shihipar claimed Anthropic intended to remove the tracker, having implemented stronger mitigations. Privacy advocates expressed concern about Anthropic's willingness to surveil users. This incident is surprising given Anthropic's refusal to allow US government surveillance of users. The tracker is seen as an aggressive measure by US firms to prevent Chinese AI firms from copying their models. Chinese firms have rapidly matched US model capabilities, even surpassing Anthropic's latest model in some areas.