Definition

Application of ML models for case classification, duplicate detection, or signal prioritization. Requires validation, version control, and explainability considerations per regulatory expectations.

Example

ML model to triage incoming reports by seriousness and completeness.

Regulatory source

EMA/EFPIA workshop on AI in PV (2025), FDA CDER AI principles

Common mistakes

Black-box models with no documentation of training data or performance metrics.

Inspector question

Describe your validation approach for any ML tools in PV.

Related terms