Threat Intelligence

AI Coding Tool Caught Saving Passwords and Sensitive Data to Public GitHub Repositories

Dark Reading · 11 Nov 2026
Key Takeaway If your business or your developers use AI coding tools, check that no credentials are being committed to public repositories, and rotate any secrets that may have been exposed.

Dark Reading reports that a command in Claude Code, an AI coding agent, can inadvertently save files containing credentials and other sensitive data into public Git repositories. The result is a security gap for organisations that use these tools in their development work.

This matters because credentials such as passwords and access keys are exactly what attackers look for. Once sensitive data lands in a public repository, anyone on the internet may be able to find it. For a small business, a single exposed credential could open the door to cloud accounts, customer data or internal systems.

The summary available does not say how many organisations are affected or which specific command is involved, so businesses should check the full report and vendor guidance for details. In the meantime, any team using AI coding assistants should review what those tools are allowed to save and where that content ends up.

AI security GitHub credential exposure Claude Code secrets management
Building or buying AI systems? Governing them under ISO 42001 ->

Summarised by CISO AI from Dark Reading, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.