AI finds unreported swimming pools for the French tax authority
More than 20,000 ‘hidden’ swimming pools have been detected since October 2021 in some of the French regions with the help of an algorithm combing through satellite photos, the French authorities announced fittingly at the end of summer 2022.
In France, there are over 3.2 million private swimming pools, and over 240,000 were built in 2021 alone (partly due to the COVID home office trends). A land with a swimming pool, however, is also subject to higher property tax rates: an average pool of 30 sqm is taxed at EUR 200 a year.
The French tax authority had partnered with Capgemini, a French consulting firm, and utilized the open-source software by Google to make sure that the additional tax is duly collected. The artificial intelligence tool deployed scans satellite images of houses and backyards, and compares those with the land registry data and flags any outliers. According to the French tax authority, the margin of error of the algorithm is small: more than 90% of owners who were identified by the tool and contacted by the authorities admitted that they did in fact have a taxable swimming pool.
The current inspection had focused on the southeast of the country and the tax office discovered 20,356 undeclared swimming pools in these areas, as a result, ‘cheques’ worth EUR 10 million were sent to unsuspecting owners. The system now being piloted is aimed to be extended to the whole country and to other house extensions (annexes, verandas) as well. Based on the initial data, 6% of the swimming pools are not registered (and thus not duly taxed), and the French tax office expects to collect additional revenues up to EUR 40 million, thanks to the advanced technology. The development and use thereof, on the other hand, allegedly cost EUR 24 million to French taxpayers already.
The initiative by the French tax office also shows a new approach by the tax authorities around the world: to make use of their vast databases combined with AI technology to find additional tax sources and predict patterns and tax risks.