French Tax Authorities Develop an AI Software That Exposes Over 20,000 Undeclared Pools

You can't run from tax authorities, especially if they are powered with artificial intelligence.

The French tax authority (often referred to as "Le Fisc") has discovered 20,356 household swimming pools that were previously undeclared using an artificial intelligence computer vision technology developed by French IT company Capgemini, according to a report by The Guardian.

The owners of the undeclared pools were given bills totaling about 10 million euros.

Detecting Pools via an Aerial Imagery

With the aid of Google's cloud computing, Capgemini's software automatically detects pools in aerial photos (by scanning for blue rectangles, for example). It compares the findings to data in real estate and tax databases to discover undeclared pools.

The owner violates tax law if it discovers that a certain address doesn't have a pool registered.

The initiative only covered nine out of the 96 metropolitan departments when it first started last October. Le Fisc reports that the system initially had a 30% mistake rate and mistook solar panels for swimming pools, but they claim that the accuracy has since improved.

French taxation on real estate is dependent on its rental value, which rises as a result of owners' construction of additions or enhancements like swimming pools.

For instance, a swimming pool that is 30 square meters in size will cost an additional 200 euros in taxes each year.

Due to the recent heat wave, private pools have recently gained popularity in France, but they are also divisive due to their water usage as they grapple with a historic drought, as noted by Ars Technica.

"We are particularly targeting house extensions like verandas, but we have to be sure that the software can find buildings with a large footprint and not the dog kennel or the children's playhouse," Antoine Magnant, the deputy director general of public finances, said in a statement with Le Parisien newspaper.

Error Margin

The technical team for the tax authorities, however, claims they are not yet able to determine if a rectangular object on an aerial image is an addition or a tent, terrace, or even tarpaulin positioned on the ground.

The software developed by Google and Capgemini was said to have a 30% error margin back in April.

The AI mistaken solar panels for swimming pools and reportedly failed to collect taxable extensions tucked away behind trees or buildings. But further tests are being conducted to make the technology better.

Magnant said that the second step of their research would allow them to verify if a property is unoccupied and no longer needs to be taxed.

There was a boom in installations before the Covid19 lockdowns of 2020 and 2021 as more people opted to work from home, and it is believed that France has since logged 3.2 million private swimming pools.

According to the public finance authority DGFiP, the AI program will now be implemented nationally and might result in an additional €40 million in private pool taxes by 2023.

This article is owned by Tech Times

Written by Joaquin Victor Tacla