Artificial intelligence is revolutionizing the fight against tax fraud by effectively tracking dishonest taxpayers. Discover in this article how these new technologies monitor even the smallest details to ensure fair taxation for everyone.
Pools, experimental ground for tax AI
Since 2022, around 140,000 owners of private in-ground pools have received a letter from the tax administration asking them to declare their pool. This letter follows the “Innovative Land” project launched by the General Directorate of Public Finance (DGFiP) two years ago. This project uses artificial intelligence (AI) to identify undeclared pools through image recognition. Pools are a good experimental ground for tax AI because they are easily detectable on maps provided by the National Institute of Geographic and Forest Information (IGN). Furthermore, pools are often rectangular, making them easier to detect. This project has allowed the state to recover up to 40 million euros so far, but it has also cost 24 million euros, resulting in a modest return on investment for the moment. The DGFiP is now considering extending this same principle to other constructions, such as garden sheds or warehouses.
Limits and challenges of using tax AI
However, the use of artificial intelligence to detect tax fraud is just beginning. The senatorial report drafted by Sylvie Vermeillet and Didier Rambaud highlights that the models used by tax AI are relatively basic and far from what is commonly done in large companies to meet their business needs. Moreover, the task is more complex for AI when it comes to detecting constructions with more heterogeneous and less colorful shapes than pools. Unions representing public finance have also reported an initial error rate of 30% from AI during the pool project. Although this rate has been reduced since then, only about ten data scientists are currently in charge of operating the machines. Compared to the IRS in the United States, which is currently hiring nearly 20,000 new profiles, some of whom are dedicated to AI, it is clear that France still has a long way to go.
Advances in AI in the fight against tax fraud
Despite these limitations, artificial intelligence has already made advances in combating tax fraud in France. Data mining, which involves cross-referencing multiple tax data, already allows for the detection of simple errors or discrepancies between professional and personal declarations. Social media is also used by the tax authorities to gather information about taxpayers and detect fraudsters. The senatorial report also suggests exploring other methods, such as unsupervised learning, which could reveal complex and unknown fraudulent behaviors by establishing statistical links between elements. The administration already possesses vast amounts of useful data to train new programs that could potentially combat fraud in other areas, such as VAT or cryptocurrencies.
Future challenges of tax AI
Nevertheless, several obstacles remain. There is no reliable estimate of tax fraud in France, complicating the assessment of the potential effectiveness of AI. Additionally, there are internal hesitations within the state, particularly in the social sphere. Finally, the administration must consider the economic consequences of using AI. If it were to become extremely effective, it might be overwhelmed by the scale of the task, which could have negative repercussions on the attractiveness of French ports.
In conclusion, the use of artificial intelligence in the fight against tax fraud in France is still in its infancy. Although progress has been made, it is important to continue investing in research and development of AI to improve its effectiveness and utilization in other areas. The collection and analysis of tax data will be essential for strengthening controls and detecting fraudulent behaviors more precisely and effectively.
