How to meet it
OWASP AISVS v1.01
Input Validation · Content & Policy Screening
Score prompts for harmful content
Verify that every prompt is scored by a content classifier for violence, self-harm, hate, and sexual content against configurable thresholds. Prompts that exceed those thresholds are rejected or sanitized before reaching the model context.
OWASP AISVS v1.01, requirement 2.2.1
In plain English
Score every prompt for harmful content such as violence, self-harm, hate and sexual content, and stop prompts that go over your thresholds before the model sees them.
What good looks like
- A content classifier runs on every prompt.
- Thresholds for each category are set deliberately and recorded.
- Blocked prompts get a clear response and are logged.
How teams usually do it
- The moderation or content safety service from your model provider or cloud platform.
- An open moderation model run in your own stack.
Evidence an auditor would accept
- Classifier configuration and thresholds.
- Test results with sample prompts in each category.
Common mistakes
- Leaving thresholds at defaults that do not match your audience.
- Screening the first message of a conversation and skipping later turns.
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.