How to meet it
OWASP AISVS v1.01
AI-Assisted Secure Coding (Appendix C) · Continuous Feedback, Adversarial Testing & Model Fine-Tuning
Red-team the AI tooling
Verify that scheduled red-team exercises target the AI tooling itself. The exercises include direct and indirect prompt-injection probes delivered through realistic PR, issue, and comment surfaces, jailbreak corpora, and supply-chain payload generation. Findings are remediated under tracked severity SLAs.
OWASP AISVS v1.01, requirement AC.6.3
In plain English
Run scheduled red-team exercises against the AI tooling itself, with fixes tracked to agreed deadlines by severity.
What good looks like
- Exercises run on a schedule.
- They include indirect injection through PRs, issues and comments, plus jailbreaks and supply-chain payloads.
- Findings have severity-based deadlines.
How teams usually do it
- Internal red team or an external firm with a scope covering AI tools.
Evidence an auditor would accept
- Exercise reports and remediation tracking.
Common mistakes
- Red-teaming the product's AI but not the development tools.
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.