Executive brief
Aider, an AI-powered pair programming tool, contains a configuration flaw where it automatically bypasses security checks (pre-commit hooks) when saving code changes to a repository. These hooks are often used by organizations to prevent the accidental upload of sensitive data like passwords or insecure code patterns. Because these protections are disabled by default, an attacker or an automated AI suggestion could introduce vulnerabilities or leak secrets into a codebase without being stopped by standard security filters.
Technical details
A vulnerability exists in Aider-AI Aider 0.86.3 due to the default configuration of the 'git-commit-verify' argument in 'aider/args.py'. The application defaults this value to 'False', which causes the 'Pre-commit Hook Handler' in 'aider/repo.py' to append the '--no-verify' flag to git commit commands. This behavior effectively bypasses repository-level security controls, including secret scanning, static analysis, and custom SDLC policy enforcement hooks. An attacker or a malicious AI-generated PR could exploit this to commit insecure code or sensitive credentials that would otherwise be blocked by the local environment's security hooks. As of the advisory date, the project has not yet released a patch to change this default behavior.
Affected products
- Aider-AI Aider 0.86.3
Timeline
- 2026-04-21: disclosed: Issue reported to the Aider-AI GitHub repository
- 2026-05-31: advisory: CVE published by VulDB/NVD