Secret Claude tracker shocks users after Anthropic’s anti-surveillance stance (arstechnica.com)

Anthropic faced backlash after a security researcher revealed that the company had secretly implemented a tracker in Claude Code that monitored users in China, raising serious privacy concerns. The hidden code utilized “prompt steganography” to discreetly gather information about users’ time zones and connections to Chinese AI labs, which Anthropic has previously accused of orchestrating distillation attacks. Although the company claimed the tracker was part of an experiment to combat account abuse and unauthorized resale of their models, its clandestine nature has been deemed a significant breach of user trust. The incident highlights a potentially troubling double standard for Anthropic, as the firm has previously resisted government attempts to use its AI for surveillance in the US. With privacy advocates criticizing the tracking mechanism, the situation raises questions about the ethical boundaries of AI companies in their efforts to protect their technologies. Anthropic's swift decision to remove the tracker reflects an acknowledgment of these concerns, but it illustrates the ongoing struggle within the AI/ML community to balance security measures with user privacy and trust.