Description
Periodic re-evaluation of AI system impacts as context, usage, or system capabilities change over time.
Implementation Guidance
Re-assessment Triggers
Define events requiring impact re-assessment: significant model updates or retraining, expansion of AI system scope or user population, regulatory changes affecting the AI use case, incidents revealing previously unidentified impacts, and scheduled periodic reviews (at minimum annually).
Re-assessment Process
Conduct structured re-assessments covering: updated risk landscape analysis, actual vs. predicted impact comparison, new stakeholder groups affected, changed regulatory requirements, and emerging risks identified since last assessment. Compare findings against original impact assessment.
Action on Findings
Document re-assessment outcomes and required actions: update risk registers, modify monitoring thresholds, adjust human oversight levels, communicate changes to stakeholders, and update system documentation. Escalate significant new risks to governance.
Evidence Requirements
- Re-assessment trigger documentation
- Completed re-assessment reports with comparison to baseline
- Updated risk registers reflecting re-assessment findings
- Action plans and completion tracking