Claude Code’s “Don’t Ask Again” File Exposes Passwords and Keys in 1,122 Public Copies on GitHub

What happened

Passwords and API keys approved by users in Anthropic’s Claude Code AI coding assistant have been found in 1,122 public GitHub repositories. The issue stems from Claude Code’s “Don’t ask again” command, which if selected, saves the exact command inputs, including sensitive secrets, into a project settings file. This allows anyone browsing those public GitHub projects to access these credentials without any special permissions.

Why it matters

Developers relying on Claude Code’s convenience feature face an unexpected security risk. By approving commands with confidential data and opting to not be prompted again, they inadvertently expose passwords and keys publicly. This significantly raises the risk of credential theft, unauthorized access, and potential breaches for affected projects. It underscores a glaring gap in how AI coding assistants handle sensitive data, weakening trust and forcing operators to re-examine automation shortcuts that prioritize ease over security. Organizations will need to audit their repositories for leaked secrets and update security protocols around AI tools.

What to watch next

Monitor how Anthropic responds to this exposure and whether they release patches or usage guidance to prevent secret leakage via Claude Code. Watch for new tooling or GitHub policies that detect and alert on AI-generated files containing plaintext credentials. Builders and security teams should track whether similar AI assistants adopt safer storage practices or tighten defaults around sensitive inputs. This incident may trigger more scrutiny on AI-assisted coding tools’ privacy practices as their adoption grows.

AI Quick Briefs Editorial Desk