What happened
Anthropic’s Claude Code AI coding assistant offers a “Yes, and don’t ask again” option that records the exact command a developer approves. When that command contains a password or API key, the tool writes the secret verbatim into a project‑settings file. Apiiro identified 1,122 public GitHub repositories containing such files, exposing dozens of credentials to the internet. The issue stems from the tool’s persistence mechanism, not a software vulnerability, and Anthropic has since disabled the feature for new projects, but existing files remain exposed.
Why it matters for trust and compliance
- The incident demonstrates a lapse in credential‑management controls and the importance of continuous repository monitoring to provide defensible evidence of secret‑handling compliance.
- Show auditors that you have controls to prevent secret storage in code and can produce evidence of secret‑scan results.
- Enable continuous monitoring of repositories to detect and remediate credential leakage in real time.
Who is affected
Technology / SaaS development teams API providers integrating AI coding assistants
Recommended actions
- Adopt a policy prohibiting secrets in code or AI‑assistant prompts.
- Implement automated secret‑scanning (pre‑commit hooks, CI pipelines).
- Rotate any exposed credentials immediately and review access logs.
- Audit AI‑tool configurations and disable persistent storage of user commands.