Skip to main content
Incidents & Monitoring
5 min readUpdated 2026-02-15

Reassessment Triggers

When and how AI systems should be reassessed — material changes, incident patterns, vendor updates, and scheduled reviews that trigger reclassification.

Reassessment Triggers

AI system classifications are not permanent. Material changes, incidents, vendor updates, and new regulatory guidance can all change a system's risk profile. Klarvo tracks reassessment triggers and prompts re-evaluation when needed.

What Triggers Reassessment

TriggerWhy It MattersAuto-Flagged?
Material system changeNew features, new model version, expanded capabilitiesYes (via change management)
Vendor/model changeNew vendor, model upgrade, provider pivotYes (vendor record update)
New use caseSystem applied to new purpose or contextManual
New affected groupsExpanded to new user populationsManual
New deployment regionExpanded to new EU member statesManual
Critical incidentSafety event, rights violation, serious harmYes (automatic)
Pattern of incidentsMultiple medium-severity incidentsRecommended
Regulatory changeNew guidance, enforcement action, or standard updateManual
Scheduled reviewPeriodic review date reachedYes (calendar-based)
FRIA update neededMaterial change affecting fundamental rights assessmentYes (linked to above)

How Reassessment Works

  • Trigger detected: Either automatically (incident, vendor change, review date) or manually flagged
  • System flagged: A "Reassessment Needed" badge appears on the AI system
  • Dashboard alert: The system appears in Compliance Alerts
  • Task created: "Complete reassessment" task assigned to Primary Owner
  • Re-evaluation: The owner revisits relevant wizard steps and updates answers
  • New classification: If answers change, the classification engine recalculates
  • Version created: A new classification version is stored (V2, V3, etc.)
  • FRIA update: If the system has a FRIA, it may also need updating
  • Reassessment vs. Full Re-classification

    Reassessment: Reviewing specific aspects that changed — faster, targeted.

    Full re-classification: Starting the wizard from scratch — thorough but time-consuming.

    Most reassessments are targeted: "The vendor upgraded the model — does this change our high-risk screening result?" You don't need to re-answer every question, just the affected ones.

    Scheduled Reviews

    Every AI system has a Next Review Date (default: 90 days after creation, adjustable):

  • Klarvo sends a reminder 14 days before the review date
  • A task is created when the date arrives
  • After review, set the next review date
  • For high-risk systems, quarterly reviews are recommended. For minimal-risk systems, annual reviews may suffice.

    Change Management Integration

    Klarvo's change management features support reassessment:

  • Material change detected → Reassessment prompt
  • Vendor record updated → Linked systems flagged
  • Classification version history → Full audit trail of changes
  • FRIA update prompt → Triggered when relevant factors change
  • Best Practices

    📅 Set review dates: Every system should have a next review date
    🔄 Don't skip reassessment: Even if you "know" nothing changed — document the review
    📝 Record the outcome: Even if reassessment confirms the same classification, record that you reviewed it
    🔍 Watch for patterns: Multiple reassessments with the same trigger may indicate a process improvement opportunity
    📋 Update FRIA alongside: If a reassessment changes the risk profile, the FRIA likely needs updating too