Prisma AIRSInstall

Summary

Performs a security scan of AI-generated responses to detect policy violations or sensitive data.

Inputs

• scanMode — Selects synchronous ('sync') for immediate 2 MB scans or asynchronous ('async') for larger up to 5 MB scans.
• content (required) — The response text to scan for threats.
• additionalOptions — Optional configuration such as AI model, profile override, application name, environment, max polling duration, retries, polling interval, timeout, transaction ID, and user ID.

Output shape

An array of execution records, each containing operation metadata, transaction ID, scan result, and optionally masked content, DLPA detection flags, and audit fields.

The node returns one record per scanned item; async scans may include additional polling fields. Errors are emitted as records when continueOnFail is enabled, otherwise they throw.

Examples

Example 1: Scan a single AI response synchronously

scanMode: sync, content: {{ $json["response"] }}, additionalOptions: {}

Example 2: Scan a large response asynchronously with defaults

scanMode: async, content: {{ $json["response"] }}, additionalOptions: { maxPollingDuration: 300000, maxRetries: 3 }

Example 3: Mask sensitive data in a prompt and scan it

operation: maskData, maskContent: {{ $json["prompt"] }}, additionalOptions: { aiProfileOverride: "production-profile" }