Anthropic Claude Code Leak Exposes 513K Lines of Source Code

A critical error in Anthropic's release pipeline exposed approximately 513,000 lines of source code for its Claude AI model. The issue occurred when a 59.8 MB source map file was inadvertently included in version 2.1.88 of the Claude Code CLI tool shipped via npm on March 31.

The exposed codebase contained internal files and unobfuscated TypeScript source, but Anthropic confirmed that no sensitive customer data or credentials were compromised. The company attributed the exposure to 'human error,' not a security breach.

A few hours after the leak, the code was mirrored to GitHub, and opportunistic actors began creating fake repositories laced with malware targeting developers who cloned the leaked source. In an attempt to address the issue, Anthropic filed DMCA takedown requests, which ended up disrupting roughly 8,100 GitHub repositories, many of which had no connection to the incident.

Crypto developers are advised to audit their dependency chains immediately and verify the provenance of any recently cloned repositories due to the malware campaigns. Additionally, reports suggest quality degradation in Claude Code responses surfaced in April 2026, further complicating the situation.