License plate

The License Plate Detection check in SUPERWISE® provides specialized computer vision safeguarding to prevent the accidental ingestion, processing, or exposure of vehicle information. This rule automatically scans image-based interactions to detect visible license plates, enabling organizations to enforce strict privacy controls and maintain regulatory compliance (such as GDPR or regional PII data protection acts).

By flagging or blocking images containing license plates, teams can protect user privacy in customer-facing applications, insurance claim workflows, automated parking assistants, or field intelligence gathering.

Configuring the rule when adding a License Plate Detection rule to your guardrail policy, you can customize the following parameters to align with your operational tolerance:

• Apply to: Choose whether this rule scans incoming visual data (Input), outgoing model responses (Output), or both.
• Confidence threshold: Specifies the minimum confidence level (ranging from 0.0 to 1.0) required for the underlying vision model to flag an object as a license plate. Setting a lower threshold increases strictness (catching more faint or angled plates), while a higher threshold reduces false positives.

Testing and Validation

Before deploying the rule to production, use the Test the rule pane on the right side of the modal. Drag and drop or upload a sample jpeg, png, or webp file (up to 512 MB) and click Run test to instantly verify how your chosen threshold reacts to actual visual data.