Anthropic Confirms Source Code Leak for Claude Code After Boast About AI-Generated Development

Anthropic confirmed parts of its internal Claude Code source code leaked after a GitHub file exposed a download link.

The company said no sensitive customer data or credentials were exposed and called the incident a packaging error caused by human error.

Anthropic issued copyright takedown requests for over 8,000 copies and the leaked code revealed proprietary details like permission enforcement and multi-agent coordination.

The leak came days after details of Anthropic’s upcoming Claude Mythos model were also leaked, raising concerns about competitive and security impacts.