The Public Finance Department has decided to generalize its tool based on artificial intelligence, developed with Google and Capgemini, to detect undeclared private swimming pools. A device tested for a year in nine departments.
The controversy largely agitated the political class during the drought episodes of this summer 2022: should private swimming pools be banned? The debate oscillated between frank provocation and legitimate questioning in a context where the consequences of global warming were more papal than ever. But for the French tax authorities, the question is approached from another angle: how to effectively tax the private swimming pools of the French? Using a computer program, the DGFIP identified more than 20,000 undeclared swimming pools, and collected several million euros in unpaid bills.
This program, called Foncier Innovant, developed with Capgemini and Google, was tested in nine departments (Alpes-Maritimes, Ardèche, Var, Bouches-du-Rhône, Rhône, Haute-Savoie, Morbihan, Maine-et-Loire and Vendée) during a year. By cross-checking satellite imagery with data from the cadastres of local authorities, the program makes it possible to detect constructions and check whether they are correctly imposed. As a result of the races, the program identified 20,356 ghost pools. The tax authorities announced on Monday August 29 that the program would be extended to the entire territory.
Other constructions soon concerned
In addition, the program will be used to identify other undeclared constructions, such as verandas, permanent pergolas or garden sheds. In short, as for swimming pools, any construction cannot be moved without destruction and increasing the land value of the property. “We are particularly targeting house extensions such as verandas, but you have to be sure that the software can find buildings with a large footprint and not the kennel or the children’s playhouse”declared Antoine Magnant, the Deputy Director General of Public Finances, to our colleagues at Parisian.
For the moment, this “artificial intelligence” would suffer from a lack of precision which could mislead it in 30% of cases. The software could thus confuse the solar panels with the swimming pools, take a cover for a construction, or not detect the presence of a building on a shaded area. However, there are very real agents in charge, one imagines, of going to do the verification firsthand If in doubt.
Profitability from the second year and doubts of the unions
The public finance authority said the scheme will now be rolled out nationwide, with a target of €40m in new taxes on private pools in 2023. “These new resources, which will be partly recurrent for local authorities, will ensure the profitability of the project from its second year of deployment”specifies the DGFIP.
For their part, the unions deplore the intervention of intelligence and data mining in these missions, which could lead to job cuts. The CGT evokes a “omerta on the intervention of Google and on data protection”but also a “hidden subcontracting of part of the operations carried out in Madagascar” and a lack of transparency on actual effectiveness and costs.