A major AI company is facing a trust crisis after it was caught secretly monitoring users. Anthropic spent three months quietly tracking individuals it suspected of connecting from China through its Claude Code developer tool. The surveillance was undisclosed and embedded in the software using a technique called "prompt steganography."
The system flagged users based on time zone settings like Asia/Shanghai and Asia/Urumqi, as well as proxy connections to Chinese domains. Anthropic stated the tracking was an experimental countermeasure against what it described as "distillation attacks," where competitors allegedly copy an AI model's capabilities.
The hidden feature was discovered by a security researcher and made public on June 30. Anthropic responded by removing the tracking mechanism within a day, releasing an update on July 1. The company stated the surveillance had been active since at least early April.
The fallout has been immediate. Chinese tech giant Alibaba moved to ban its employees from using Claude Code, with the restriction taking effect on July 10.