Zscaler AI Guard icon

Zscaler AI Guard

Scan AI prompts and responses for security threats using Zscaler AI Guard

Actions3

Dual Scan

AI-generated

Summary

Scans both an AI prompt and its response for security threats, stopping the response scan if the prompt is blocked.

Inputs

  • promptContent (required) — The prompt content to scan.
  • responseContent (required) — The response content to scan.
  • additionalOptions — A collection of optional configuration settings including AI model identifier, application name, environment, timeout, retries, policy ID override, custom transaction ID, and user ID.

Output shape

A single item containing the scan result with metadata, including the operation type, severity, detected and blocking detector names, a boolean blocked flag, and workflow context details.

If the prompt scan results in a BLOCK action, the operation returns the blocking result immediately and skips the response scan. If the response scan is performed, the result includes the response scan data alongside the prompt scan result. Content exceeding 5MB is rejected.

Examples

Example 1: Blocking Threat Detection

Inputs: Prompt 'Generate a phishing email', Response 'Here is the email...', Options: Default

Example 2: Pass-Through Scan

Inputs: Prompt 'Hello', Response 'Hi there!', Options: Environment set to 'production'

Example 3: Long Content Handling

Inputs: Large prompt and response content, Options: Timeout increased to 60,000ms

Links

Discussion