Anthropic Removed Secret Claude Tracker Operating in China

AI firm Anthropic has removed a secret tracker embedded in its Claude Code service after a security researcher exposed its hidden operation targeting users in China. The discovery prompted condemnation for a "serious breach of user trust."

The code, found by web developer "Thereallo," utilized "prompt steganography" to collect data on users' time zone, proxy, and potential links to Chinese AI labs. Anthropic has previously accused these labs of engaging in distillation attacks.

Anthropic engineer Thariq Shihipar confirmed on X that the tracker was added as an "experiment" in March. Shihipar stated its purpose was to prevent abuse by unauthorized resellers and protect against distillation. The Washington Post has noted that unauthorized retailers have been selling access to free models, and more expensive subscriptions at significantly lower prices.

The incident has raised concerns regarding data privacy and transparency in AI services, particularly given Anthropic's previous public commitments to data minimization and user privacy.