Skip to content
CERT Foundation
AI-assisted review

Machines screen. Named humans decide.

CERT uses machine learning to screen submissions and MRV data for anomalies, duplicates and over-crediting risk. Every flag is reviewed by a named human accountable for the decision.

What the models do

  • Anomaly detection on monitoring time-series.
  • Cross-registry duplicate detection using coordinates, activity data and identifiers.
  • Over-crediting risk scoring against sectoral distributions.
  • Text-analysis on PDDs and verification reports to flag inconsistencies.
Machines do not decide
No AI flag becomes a decision until a named human reviewer signs off. Every decision — for or against — is documented and published.

Model governance

Model architecture, training data classes and performance metrics are published. External model audits are commissioned annually.