Skip to main content

Guardrail Violations

.

Guardrail Violations

If your organization has enabled Guardrails, use the Guardrail Violations page to monitor blocked queries and assess the effectiveness of the content filter.

biggy_webapp_guardrailviolations.png

Use the drop-down menu at the top right of the screen to adjust the selected time period.

The following report widgets are available in the Guardrail Violations section:

Widget

Description

Total Violations

Number of queries that were blocked by guardrails.

Blocked Today

Number of queries that were blocked by guardrails today.

Top Violation

The guardrail policy that has the highest number of policy violations associated with it.

Top Offending User

The user who has the highest number of guardrail violations.

Daily Violations Trend

A graph displaying the number of queries blocked by guardrails, over time. Hover over a specific day on the graph to view the number of blocked queries on that date.

Violation Category Breakdown

The number and percentage of violations by each high-level guardrail category. (Core safety, Custom, Optional)

Top Offending Users

List of users with the most guardrail violations.

Hover over a user to view their total number of violations and their most common guardrail violation category.

Top Violation Reasons

The most common reasons guardrails were triggered, and the number of times each reason occurred.

All Violations

Detailed records of all queries blocked by guardrails. The following information about each blocked query is available in the list:

  • Date/Time - date and time the query occurred

  • User - user who attempted the query

  • User query preview - shortened text of the attempted query

  • Type - high-level category of guardrail that was triggered

  • Reason - reason the guardrail was triggered

  • Platform - platform the violation took place on

Click any query in the list to view additional information. The following appears in the full details:

  • Full user query - full text of the query

  • User message shown - Biggy's response to the query violation

  • Detailed reasoning - full details behind why the query was blocked by a guardrail

  • Category - guardrail type

  • Severity - severity level of the violation

  • Request ID - unique ID of the query

You can filter the list by User, Type of violation, Reason for the violation, or Platform (Slack, Teams, or Biggy Web Chat).

To view the report in CSV format, click Export to CSV.