Anthropic's AI Tool "Claude Code" Exposes Code Due to Source Map Mismixing

Claude Code's source became accessible after the public npm package v2.1.88 accidentally included source map files.

The exposure spread on March 31 (local time).

Anthropic says this was not a security breach but a human packaging error and that it is implementing measures to prevent recurrence.

The code was shared on GitHub and spread in the developer community after being pointed out on X; some developers are analyzing memory structures.

Anthropic states no customer secrets or credentials have been confirmed leaked, and some developers are reimplementing the tool from scratch (e.g., in Python).