Researchers expose GitHub's AI agent can leak private repositories

Security researchers demonstrated a vulnerability where GitHub's AI agent could be manipulated into exposing access to private repositories. The attack highlights a real risk as AI agents gain deeper integration with development tools—makers relying on these tools for sensitive code should review their repository permissions and AI feature usage.