We’ve all heard the warnings that machines are coming for our jobs — but in France, they’re first coming for our pools.
The country’s tax authorities have successfully trialed AI that can help them find undeclared swimming pools, and tax them. So far, according to The Guardian, this piece of software has brought in around €10m in bills for the government across nine French departments (roughly equivalent to the US’ counties).
Needless to say, nobody likes paying extra taxes. But such a development does showcase how far AI has come in processing large amounts of data and interpreting it despite the large number of variables involved. Over time, the system should also become able to notice other undeclared extensions to properties such as annexes, verandas, or permanent pergolas. That being said, it will also monitor for properties that are no longer in use and, as such, shouldn’t be taxed.
No taxation without monitoring
“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,” says Antoine Magnant, the deputy director general of public finances, for local media. “This is our second stage of research and will also allow us to verify if a property is empty and should no longer be taxed.”
The system was developed by Google and Paris-based IT company Capgemini. Its role is to analyze aerial images to determine the layout and elements of various properties and then compare its findings with data available in land registries.
Since property taxes in France are calculated based on the rental value of each property, improvements such as swimming pools can mean an increase is due. Such improvements must be declared to the tax office within 90 days of completion. In very broad lines, the team behind the AI explains, a 30 square meter pool could lead to an increase of €200 every year in taxes.
But if these improvements are not declared to officials… well, then there’s no increase in tax.
Manually monitoring all properties to ensure that all buildings or improvements they contain are properly declared is simply not feasible. But an AI could perform this task in a cost-effective manner. French authorities decided to see if one such system can be made to spot undeclared annexes. The answer seems to be ‘Yes’.
Since its launch one year ago, the AI has uncovered 20,356 pools across the nine departments it was tested on.
Given the good performance of the system, French authorities will expand its use to the rest of the country. They also hope to expand its abilities so that it can spot other kinds of undeclared annexes or extensions, including verandas and permanent pergolas.
However, right now, that is beyond the AI’s reach. It is not yet able to determine whether a rectangular shape seen from an aerial image is, for example, a tent, a terrace, an extension to a building, or a simple tarp spread on the ground. But it’s still come a long way since its early days. Back in April, the software was claimed to have a 30% margin of error and was mistakenly identifying solar panels as swimming pools; it also consistently failed to identify taxable extensions if they were covered by trees or shadows. Since then, it’s come a long way and although it’s not perfect yet, the tax authority’s technical team is working on improving the technology.
Once perfected and rolled out across the whole of France, the system could lead to an increase of €40m in new taxes on private pools in 2023 alone. The country is estimated to have 3.2 million private swimming pools, many of which were constructed after 2020 — but not all of them declared.
Environmentalist groups across France are also applauding the initiative as they hope that the system can lower water use and water waste in France by making it harder for people to dodge taxes on swimming pools. This is especially important during the current drought, they argue, when drinking water supplies are genuinely threatened.