@xiangxiang103: Wow, Anthropic really dropped the ball this time. Someone dug up hidden code in the Claude Code binary — specifically designed to detect whether you're a Chinese user or routing through China. Not ordinary telemetry, but deliberately obfuscated, not mentioned in release notes, and completely unknown to users. The process goes like this: - Detects you...

Wow, Anthropic really messed up big this time. Someone dug up hidden code in the Claude Code binary — specifically designed to detect whether you're a Chinese user or routing through China. Not ordinary telemetry, but deliberately obfuscated, not mentioned in release notes, and completely unknown to users. The process is quite devious: - Detects you're using a proxy → checks the timezone (Shanghai/Urumqi) - Checks if the proxy URL points to Chinese domains or AI labs - If triggered, secretly alters characters in the system prompt: changes date formats, replaces apostrophes with other Unicode homoglyphs - These changes are invisible to the naked eye, but Anthropic's backend can match them In short: it embeds a set of watermarks in your prompt that only they can decode. It's been there since v2.1.91 (April 2 this year). The code is XOR encrypted, function names are minified into a mess — strings scanning can't find them. Everyone knows what Anthropic is trying to prevent: API reselling for bypass services, model distillation by domestic labs. But here's the problem — you're building a developer tool. Users give you file system access, they give you Shell. And you pull this behind their backs? What hurts most: zero documentation, zero disclosure. If they can secretly watermark your prompts today, what about tomorrow? The day after? A coding agent running on your local machine — you never know what else it's hiding. The source code was leaked via npm back in March, and 600,000 lines were turned inside out. Now, not long after, another incident. The bottom line for developer tools is three words: don't stab in the back. Trust takes years to build, but only one sneaky move to shatter. This is already exploding on X and Reddit. I don't believe Anthropic can play dead on this. Original post link: