AI system
- System
-
Foncier innovant (DGFiP deep-learning detection of pools and buildings on IGN aerial photographs, built with Capgemini and Google)
- Model
-
Not assessed
- Release
-
Not assessed
- Application developer
-
Direction générale des Finances publiques (DGFiP) with Capgemini and 
Google - Role
-
Not assessed
- Configuration
-
Not assessed
- Inputs and outputs
-
Not assessed
- Autonomy
-
Not assessed
- Workflow
-
IGN aerial photographs covering the whole territory. A detection model built with Capgemini and Google, generalized nationally after the 2022 nine-département experiment. Comparison of detected outlines with cadastral declarations. Agent verification of each detection and letters inviting owners to regularize.
- Human oversight
-
Not assessed
- Relation to AI
-
The AI detects pool outlines on aerial photographs and compares them with declarations. An agent verifies each detection before the construction is added to the cadastral value (Ouest-France). The 140,000 pools and €40 million are the minister's own attribution to the tool, relayed by several news outlets. CNews cites the DGFiP annual report via Ouest-France. No official document is among the sources. A CGT Finances publiques official told Ouest-France that field colleagues put the algorithm's error threshold at 55–60%. So the detections required substantial human checking. No commune-level outcome is on record. The Dordogne figure is the most local one read.
- AI contribution
-
The AI detects pool outlines on aerial photographs and compares them with declarations. An agent verifies each detection before the construction is added to the cadastral value (Ouest-France). The 140,000 pools and €40 million are the minister's own attribution to the tool, relayed by several news outlets. CNews cites the DGFiP annual report via Ouest-France. No official document is among the sources. A CGT Finances publiques official told Ouest-France that field colleagues put the algorithm's error threshold at 55–60%. So the detections required substantial human checking. No commune-level outcome is on record. The Dordogne figure is the most local one read.
Benefits
- Outcome identifiers
-
Not assessed
- Benefit
-
detected 140,000 undeclared pools, c. €40m extra property tax
- Category
-
Increased Product Revenue
- Status
-
Not assessed
- Beneficiary groups
-
People. Organizations. Society.
- Beneficiaries
-
Not assessed
- Reach
-
Not assessed
- Distribution and access
-
Not assessed
- Duration
-
Not assessed
- Expected?
-
Yes
- Costs and harms
-
Not assessed
Measurement
- Studies
-
Not assessed
- Study design
-
Not assessed
- Measurement
-
Not assessed
- Setting and sample
-
Not assessed
- Measures
-
detected 140,000 undeclared pools, c. €40m extra property tax
- Comparison
-
Assumed alternative: a surveyor noticing a new pool in a neighborhood by eye and checking the cadastre
- Results
-
Assumed comparison. Reference: le géomètre qui remarquait une nouvelle piscine dans un quartier et devait vérifier dans le cadastre si elle était déclarée ou pas. Result: Dans toute la France, 140.000 piscines clandestines ont été repérées, ce qui va rapporter 40 millions d'euros.
- Additional value from AI
-
Not assessed
- Limitations
-
Not assessed
- Denominators
-
Not assessed
- Materials for replication
-
Not assessed
Money
- Kind of value
-
additional property tax (taxe foncière) to be collected on the 140,000 detected pools
- Amount
-
Estimate: 40 millions d’euros
- Currency and prices
-
Not assessed
- Period and unit
-
Not assessed
- Realization
-
Not assessed
- Recipients and payers
-
local authorities (collectivités), which receive the taxe foncière
- Gross or net
-
Not assessed
- Included costs
-
Not assessed
- Comparison
-
Not assessed
- Method
-
Not assessed
- Return and payback
-
Not assessed
People and places
- Participants
-
Developer:
Direction générale des Finances publiques (DGFiP) with Capgemini and Google. Operator: Direction générale des Finances publiques (DGFiP), ministère de l'Économie et des Finances. - Jurisdictions
-
Benefit:
France (national; Dordogne example). Developer: France. Operator: France. - Ownership and control
-
Not assessed
- Dependencies
-
Not assessed
- Cross-border connections
-
Not assessed
- Geopolitical implications
-
Not assessed
Follow-up
- Conditions for replication
-
Recent, high-resolution aerial coverage. A false-detection rate low enough for agents to verify (a union official cites 55–60% errors). Legal basis to add undeclared constructions to the cadastral value. Local offices with capacity to send and handle catch-up notices.
- Replication attempts
-
3 follow-up entries. Benefit recurred: 2. Outcome unresolved: 1.
- Later findings
-
12 June 2024. In June 2024 the DGFiP's 2023 activity report, as relayed by 20 Minutes and CNews, restated the 140,000 pools and the 40 million euros of property tax, against 20,000 pools in 2022, and the plan to extend detection to undeclared buildings.
- Availability
-
Not assessed
- Institutional decision
-
Not assessed
- Grounds
-
Not assessed
- Action and result
-
Not assessed
- Success criterion
-
Not assessed
Evidence
- Source numbers
-
AIBD-0335-source-0001–0006
- Citations
-
Numbered references below.
- Language
-
Not assessed
- Source relationships
-
Independent
- Original evidence
-
Not assessed
- Access and versions
-
Not assessed
- Supported findings
-
Not assessed
- Evidence links
-
Not assessed
- Evidence strength
-
Not assessed
- Missing evidence
-
Not assessed
Grades
- Assessed benefits
-
People. Organizations. Society.
- Depth
-
People: Not assessed. Organizations: 2. Society: 1.
- Reasons
-
People: Owners of detected pools bear a cost; no individual beneficiary is documented. Organizations: Material to the DGFiP and to local authorities. The minister attributed about €40 million of additional property tax to the tool. Dordogne alone is recovering tax on about 1,500 pools. Society: About €40 million nationally is small against total property-tax receipts. The press accounts disagree on the yearly split.
- Observed or intended
-
Not assessed
- Rubric
-
Not assessed
- Assessor and date
-
Not assessed
- Alternative assessments
-
Not assessed
- Fill
-
Independently reported
Entry details
- Kind
-
Not assessed
- Related entries
-
Not assessed
- Field
-
Public finance: property-tax base and anti-fraud. Tax administration.
- Purpose
-
Detect undeclared pools and buildings from aerial imagery so that they are added to the cadastral base and taxed. This was the tool's intended purpose.
- Dates
-
Date basis unassessed: 2024 (precision unassessed). Updated: 21 October 2026.
- Stage
-
Not assessed
- Version
-
SHA-256: 87e51983bab4…
References
- Other identifiers
-
Not assessed
- Category mappings
-
Not assessed
- Brief
-
Not assessed
- Revisions
-
8 revisions. Latest: 21 October 2026.
Sources
-
News report Independent author
Dussourt, G. (2024). Grâce à l'IA, l'Etat découvre 140.000 piscines non déclarées et récupère 40 millions d'euros. RMC, 21 Mar 2024 (quoting minister Thomas Cazenave).
ce sont 140.000 piscines qui ont été détectées et qui permettront à nos collectivités de récolter 40 millions d’euros au titre de la taxe foncière
-
News report Independent author
Ouest-France (2024). Garages, vérandas, piscines… Comment le fisc repère les bâtiments non déclarés. 21 Mar 2024.
l’outil algorithmique introduit par la DGFiP en octobre 2021 avait permis la détection de 140 000 piscines, pour 40 millions d’euros de recettes supplémentaires
D’après mes collègues sur le terrain, le seuil d’erreurs des algorithmes atteint 55 à 60% en ce moment
-
News report Independent author
CNews (2024). Piscines non déclarées : le fisc repère 140.000 fraudeurs grâce à l'IA. 11 Jun 2024 (citing the DGFiP annual activity report via Ouest-France).
En 2023, cette technique a permis à la Direction générale des finances publiques d'identifier 140.000 piscines non déclarées, selon son rapport annuel d'activité publié par Ouest France.
-
News report Independent author
Bertrand, M. (2024). Le fisc découvre 1.500 piscines non-déclarées en Dordogne grâce à l'intelligence artificielle. ici (France Bleu Périgord), 24 Mar 2024.
Les finances publiques de Dordogne ont repéré quelque 1.500 piscines non déclarées au fisc en Dordogne, au cours de l'année 2023, sur les quelque 30.000 piscines du département.
-
News report Independent author
20 Minutes (2024). Taxe foncière : Grâce à l'IA, le fisc a repéré 140.000 piscines non déclarées en 2023. 12 Jun 2024.
En 2022, 20.000 piscines non déclarées avaient ainsi été repérées pour un redressement estimé à 10 millions d’euros.