France has been quietly building one of the most advanced AI-powered tax systems in the world. What started as a simple fraud detection tool has grown into a vast surveillance network and taxpayers are increasingly finding that there is nowhere to hide.
The French Tax Administration (DGFiP) began using AI as far back as 2014, with the introduction of a system known as the CFVR, a tool designed to target fraud and process large volumes of data automatically. At the time, it was a modest step. Tax officials were looking for a faster way to flag suspicious returns without relying entirely on manual checks.
But the real turning point came in the late 2010s. The administration began building what it calls a "data lake", a massive database combining tax records, banking data, real estate information, and social data. Predictive algorithms were then trained to assign risk scores to individual taxpayers, identifying who was most likely to be under-reporting income or hiding assets.
The results were striking. By 2022, the system was generating around 155,000 audit proposals per year, three times more than in 2018. That same year, the DGFiP issued €14.6 billion in tax reassessments.
One of the more surprising applications of this technology has been in property tax. Using aerial photography and machine learning, French tax authorities began scanning images of residential properties to identify undeclared improvements such as swimming pools, verandas, sheds, and extensions that owners had failed to report.
Under French law, any significant upgrade to a property must be declared to the tax authorities, as it affects the property's assessed value and, consequently, the taxes owed. Many homeowners had simply never bothered to report these changes.
The AI caught them. In one department alone, tax authorities recovered more than €200 million in a single year through audits triggered by this imagery analysis. The Alpes-Maritimes region, where swimming pools are particularly common due to the warm climate, was one of the most productive areas for this kind of enforcement.
In January 2025, France took a significant new step. A decree signed on December 31, 2024 gave the DGFiP permission to monitor the public social media activity of taxpayers using automated tools and AI.
Tax officials can now create their own accounts and browse publicly available content on platforms like Facebook, Instagram, TikTok, and LinkedIn. They are looking for two things in particular: signs that a taxpayer's lifestyle does not match what they have declared on their tax return, and evidence that someone is living in France while claiming for tax purposes to be a resident elsewhere.
This kind of check had been difficult to carry out before at scale. Now, it is increasingly routine.
The move is part of a broader shift in how France approaches tax enforcement. Rather than auditing taxpayers at fixed intervals, the DGFiP now targets individuals based on algorithmic risk scores. If your data profile raises a flag, an audit is likely to follow.
Also introduced in early 2025 was an extension of the statute of limitations for false residency cases. Previously, the DGFiP had a standard window to investigate. Now, for cases involving taxpayers who falsely claim to live outside France, that window has been extended to 10 years.
For high-net-worth individuals, expatriates, and non-residents with assets or ties in France, the message from tax authorities is clear: the old strategies of opacity and delay are no longer reliable.
Despite the scale and sophistication of these tools, France has maintained one important safeguard: a human is always involved before any formal action is taken. The AI identifies and flags; a tax official reviews and decides.
France is not alone in this approach, the Netherlands, Belgium, and other EU countries are also using AI for tax compliance. But France is widely seen as a leader in the field, and its model is being studied closely by other governments.
For ordinary taxpayers, the practical implication is straightforward: the chances of an audit being triggered by an undeclared pool, an inconsistent tax return, or a lifestyle that does not match declared income have never been higher. The system is not perfect, but it is fast, scalable, and getting smarter every year.
