AI Guardrails categorizes prompts and responses into AI threat and inappropriate content categories. However, because detection accuracy depends on your policy configuration, you must tailor it for optimal results.
Before deploying your AI Guardrails policies to your organization, you can test sample prompts against the AI Guardrails profiles you’ve configured in the AI Guardrails sandbox to confirm it detects and classifies the traffic as expected. This lets you push your policies live with confidence, knowing it will perform as intended.
To test prompts in the AI Guardrails sandbox:
-
Go to Policies > AI Guardrails.
-
Click the Sandbox tab.
-
Select the AI Guardrails profiles you want to test prompts for.
-
Enter the prompt you want to test in the text box.
Netskope let’s you know if it matched a category in any of the AI Guardrails profiles you selected:
At anytime, you can:
-
: Click to copy the results of your test, such as the profiles and categories matched. -
: Click to retry and test the prompt again, especially useful if you’ve gone back and edited the AI Guardrails profile you’re testing.