AI-Assisted Secure Coding (Appendix C) · AI Code-Review & Assistant Bot Hardening
A policy engine authorises bot actions
Verify that any privileged action a bot can take (approving a PR, merging, labeling, dismissing reviews, posting comments outside its sandbox, invoking external tools) goes through a separate, audited authorization path. That path is adjudicated by a policy engine, not by the LLM.
OWASP AISVS v1.01, requirement AC.11.5
In plain English
Any privileged action a bot can take, such as approving, merging, labelling, dismissing reviews or calling external tools, goes through a separate, audited authorisation path decided by a policy engine.
What good looks like
- Privileged actions are routed through a policy engine.
- The model's output cannot authorise itself.
- Decisions are logged.
How teams usually do it
- A small service between the bot and the repository API that applies policy.
Evidence an auditor would accept
- Policy rules and decision logs.
Common mistakes
- Giving the bot's token write access and trusting the model to behave.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.