Thousands of sensitive secrets including passwords and API tokens leaked across open source projects due to configuration storage practices in the Claude Code AI development tool. The exposure occurs when configuration settings automatically record shell commands alongside security data.

Key Developments and Technical Context

Cybersecurity researchers at Apiiro discovered thousands of secrets including database passwords, API tokens, and root logins inside Claude Code settings files hosted on GitHub.

The leak stems from the way the software handles user approval mechanisms for automated shell commands.

When a user permanently approves a command, that choice alongside any sensitive information embedded within it gets saved into the .claude/settings.local.json file on the personal computer.

Anthropic documentation mandates adding this settings file to the global gitignore list to prevent public sharing.

Strategic Analysis and Industry Implications

This incident illustrates the risks when AI-assisted coding tools automate local workflows without clear warnings that sensitive data might be captured within log or configuration files.

Outlook and Key Factors to Watch

The exact scope of affected repositories and the total number of developers who unintentionally pushed these credentials to public environments remain fully unquantified.