Vulnerability
Wrong suspension can immediately harm a household.
Tax reform · Plan de Rupture
6.10 — Fight social-benefit fraud with artificial intelligence
Detection must target intentional fraud without turning a reporting mistake, life change or incomplete data into automatic guilt.
What the Plan proposes
Plan source: Recette nette — chiffrage détaillé à expertiser. The title and stated effect are preserved from the public edition; this page adds an initial legal and operational architecture for expert review.
Detection must target intentional fraud without turning a reporting mistake, life change or incomplete data into automatic guilt.
Status: structured civic prototype requiring expert review before filing or implementation.
Current system
The ministry reported €3 billion in social fraud detected and reassessed in 2025. Final reporting must distinguish detection, notification, successful appeals and actual collection.
Implementation path
- Define fraudSeparate intent, error, overpayment and unreported change.
- Link data lawfullyDocument each source, authority and retention period.
- Create a right to explanationState facts and documents, not merely a score.
- Protect continuity of rightsAvoid automatic suspension when facts are disputed.
- Measure net recoveryDeduct reversals, cost and uncollectible debt.
Key risks
Targeting bias
The most checked groups may appear the most fraudulent.
Stale data
Outdated records may create false suspicion.
Appeal
Correction must be rapid and traceable.
Indicators to publish
| Indicator | Before | After | Purpose |
|---|---|---|---|
| Scope and population | Baseline | Annual value | Prevent perimeter changes |
| Budget effect | Estimate | Cash collected or spent | Separate promise from execution |
| Transition cost | Disclosed | Multi-year tracking | Calculate net effect |
| Appeals and errors | Initial rate | Corrected rate | Protect rights |
Official sources
consolidation audit
Category 11 financial audit — social support, transparency and anti-corruption →