Anthropic's Claude Code tracks API requests with hidden Unicode, sparking backlash.

Anthropic's Claude Code has sparked significant backlash on Hacker News after developers discovered it uses steganographic markers to silently track and classify API requests, raising alarms about privacy and unauthorized API monitoring. Security analysis reveals the software embeds hidden Unicode characters into system prompts by substituting standard ASCII with visually identical Unicode alternatives, activating only when a custom API base URL is configured to detect proxy resellers, unauthorized distributions, and potential model distillation attacks. The revelation has fueled anger among developers concerned that their prompts are being covertly altered without consent, while critics warn that these steganographic signatures could eventually leak into legitimate training data, complicating future AI development and eroding trust in the platform's transparency.