Unintentional Exposure of Claude Code’s Source by Anthropic in npm Packaging Mishap
Anthropic confirmed on March 31, 2026, that it inadvertently exposed the complete source code of its popular AI coding assistant, Claude Code, due to a packaging error when publishing version 2.1.88 to the public npm registry. The leak exposed approximately 512,000 lines of TypeScript code across nearly 2,000 files.
The incident occurred when a 59.8 MB source map file, intended for internal debugging purposes, was accidentally included in the public package release. Developer Chaofan Shou first discovered and publicly flagged the leak on social media, and within hours the codebase was mirrored across GitHub and analyzed by thousands of developers worldwide.
The leaked code revealed Claude Code’s internal architecture, unreleased features, and development roadmap, including capabilities like “self-healing memory,” persistent background assistants, and autonomous daemon modes that were fully built but not yet shipped to users. Anthropic quickly removed the package from npm, but the code had already been widely distributed.
“This was a release packaging issue caused by human error, not a security breach,” an Anthropic spokesperson stated, adding that no sensitive customer data or credentials were exposed. The company is implementing measures to prevent future occurrences. This marks the second major security lapse for Anthropic in recent weeks, raising questions about operational security as the company prepares for a potential IPO. The incident occurred as Claude Code revenue had reached a $2.5 billion annual run-rate as of February 2026.
Source: Axios