January 20, 2026 WitnessAI Update

This release adds support to additional Guardrails to the recently released Alert Severity Customization system, granting policy authors control over alert severity levels.

This change addresses the key customer problem of all Alerts being set to ‘High’ severity, which previously made prioritization and triage difficult.

New Default Alert Severities of High, Medium, and Low have been applied to all Guardrails as a starting point for your organization.

Key Capabilities

Customizable Alert Severity

Policy authors now have the flexibility to set Alert Severity to High, Medium, or Low for GuardRail Rules, as well as Data Types in the Guardrails where they appear.

New Default Tiered Severity

Default Alert Severities have been established as High, Medium, and Low, and have been applied to all the supported Guardrails listed below.

Supported Guardrails for this Release

Behavioral Activity

Data Protection

Harmful Response Prevention

Model Identity Protection

Model Protection

Risk Analysis

Support for the Organizational Behavior Guardrail is expected to be released shortly.

Dependencies and Related Changes

To support custom severity settings for Data Protection, Policy authors can now toggle individual Data Types on/off in the Guardrail configuration.

Model Identity Protection will no longer require the user to provide the System Prompt or Select a model.

Feature Updates

Alert Console

Enhanced Alert Console Triage: Console users can now efficiently review and prioritize Alerts on the Alerts page with new filtering capabilities for Alert Severity:

Ability to filter Alerts by High, Medium, and Low severity.

Alert Severity is now displayed in the Table view and the Alert side panel.

Known Issues

No known issues identified in this release.

Fixed Issues

No fixed issues in this release.

Breaking Changes

No breaking changes identified in this release.

Please contact Customer Success if you have questions or need support with this release.