Massive AI Code Leak Exposes Critical Security Flaws

On March 31, 2026, Anthropic unintentionally published the full source code of Claude Code to the public npm registry. The leak involved roughly 512,000 lines of TypeScript across 1,906 files, including concealed feature flags and mentions of an unreleased AI model named Mythos. This sensitive data was left accessible on a Cloudflare storage bucket until a security researcher discovered it and brought it to light.