France's Digital Inquisition

Summary

Joint investigation with Le Monde exposing CNAF's decade-long use of machine learning to score nearly half of France's population for welfare fraud risk. Found the system directly and indirectly discriminates against groups protected under French discrimination law. Revealed CNAF had never audited its model for bias. Documented intrusive home inspections including unannounced visits and counting household items. Reported that seven of ten investigations are algorithm-flagged.