Anthropic's Worst Nightmare — The Claude Code Source Code Leak скачать в хорошем качестве

Anthropic's Worst Nightmare — The Claude Code Source Code Leak

512,000 lines of code. 1,906 internal files. One accidental npm publish. And Anthropic's "safety-first" reputation in freefall.On March 31, a debugging source map was accidentally bundled into Claude Code v2.1.88 on the public npm registry. Within hours, the entire codebase was mirrored across GitHub — and what developers found inside was way more explosive than anyone expected.Undercover mode telling Claude to never reveal it's an AI when making open source commits. Frustration tracking scanning every message you type for curse words and negativity. Anti-distillation systems injecting fake tools to poison competitor training data. A secret background agent called KAIROS that runs without you even knowing. And 44 fully-built features hidden behind feature flags that Anthropic never told anyone about.Then came the cleanup: 8,100 GitHub repos hit with DMCA takedowns — most of them innocent forks of Anthropic's own public repo. Anthropic called it an "accident." The developer community called it something else entirely.This is the second major data blunder from Anthropic in under a week. The company that sells itself as the most responsible AI lab just became its own biggest security liability.We break down everything: how the leak happened, what was found, why it matters, and what it means for anyone using Claude Code right now.Subscribe to ContentBuffer for deep-dive tech analysis every week. Drop a comment — should Anthropic have open-sourced Claude Code instead?www.contentbuffer.com