The analysis

New technologies and old doubts

Operation foreseen by the 2025 manoeuvre kicks off: 65 million euro recovery expected Trade associations call for attention to anomalies, from distributors to group lunches

by Salvatore Padula

12 January 2026

Aggiungi Il Sole 24 Ore
ai preferiti su Google

4' min read

Translated by AI

Versione italiana

4' min read

Translated by AI

Versione italiana

The tax delegation law announced a major commitment on the fight against tax evasion. With the watchword of a 'paradigm shift', it guaranteed a gradual shift in the fight against tax evasion from ex post assessment to ex ante tax risk prevention and selection models. It also promised - as was later envisaged in one of the implementing legislative decrees (13/2024) - to improve, strengthen and rationalise the techniques of tax risk analysis with the use of new technologies: algorithms, artificial intelligence, machine learning and other advanced data analysis techniques.

At the Origins of Risk Analysis

To tell the truth, the focus on risk analysis techniques was not born with the delegation. There is already a trace of it in 2011 (Article 11(4) of Decree-Law 201) and more recently in the Budget Law for the year 2020 (160/2019), which allowed experiments to begin. On the basis of these regulations, for example, the algorithm was created that correlates the archive of financial reports with other databases used by the administration. This is Vera (acronym for verification of financial relations), an algorithm regulated by a decree of the Economy in June 2022, after accepting the observations of the Privacy Guarantor. Guarantees that today represent a point of reference.

Of course, the knot of the balance between public interest and the protection of taxpayers' rights remains a central and very delicate element, as confirmed, albeit on similar but not identical ground, by the recent decision of the European Court of Human Rights (40607/19 and 34583/20) on access to bank accounts, which raises many questions on the discretionary nature of the IRS's actions.

In any case, the lines of action on taxation and technology have existed since before the enabling act. However, the enabling act and its implementing decree are considered fundamental because, as Alessandro Santoro wrote in lavoce.info, they define and strengthen the legal framework of risk analysis. They identify its basic elements. And, moreover, they extend the analysis to the data in all the archives to which the administration has access (with a few exceptions).

The Evasion Numbers

Meanwhile, the country continues to reckon with ever alarming levels of tax evasion. The figures are improving, at least so we read in the 2025 report on tax evasion, yet in 2022 (the last year surveyed) a mountain of money worth between 89.7 and 90.9 billion euro (up to over 102 billion if social security contributions are included) has 'disappeared'. It is, of course, a good thing that the propensity to evade is reduced to around 17%, also thanks to the measures introduced in past years. But the critical points persist: and it is a fact that the IRPEF evasion of VAT holders (businesses and professionals) continues to fluctuate at very high levels and has reached the absolute value of a good 36.9 billion euro.

In short, it seems to be understood that some mechanical control and recovery tools (electronic invoice and split payment), combined with some innovative elements gradually introduced by the Revenue Agency - from machine learning techniques to network analysis, as Director Vincenzo Carbone recalled in a parliamentary hearing - have worked well, especially on VAT and where the database is more solid. But the control tools for smaller taxpayers remain weak.

The point is that the breakthrough still remains in limbo. Because the concrete implementation of the provisions on risk analysis contained in Legislative Decree 13/2024 is suspended, pending the necessary implementing regulation by the Minister of the Economy.

Strategies for action and positive signals

Of course, the absence of the provision does not determine the interruption of innovative selection and control activities, which continue on the basis of the previous provisions. The administration, it must be said, is not standing still. The agencies (Revenue and Collection) together with the Gdf have set up the Integrated Risk Analysis Unit, a coordination structure consistent with the framework outlined by Legislative Decree 13/2024 on risk analysis. But it is clear that without the implementing rules and opinions of the Privacy Authority, the full implementation of the advanced risk analysis model outlined by the reform will be slowed down and the full potential of artificial intelligence will not be exploited.

In the meantime, however, there are encouraging signs. Since the beginning of the year, the administration has been able to count on a few new tools for countering and deterring. They range from the start of the 'running-in' of the procedure that imposes the pairing of Pos and cash register to the other regulations envisaged by the Budget Law that, in addition to aiming at speeding up compulsory collection, provide for a further strengthening of risk analysis activities.

In conclusion, it is the usual Italian paradox. On the one hand, we have a regulatory framework that emphasises the use of digital tools to track down evaders. On the other, delays in the implementation phase and (legitimate) prudences related to the protection of privacy end up weakening the reform. One almost gets the impression of a system that is content to communicate a great commitment against evasion.

We shall see what happens. Meanwhile, with the end of the legislature approaching, the political system will soon enter a permanent campaign climate. And it is legitimate to ask oneself how politically feasible it is, in a phase such as the one that is approaching, to make fully operational a model for combating tax evasion that, by its very nature, is likely to be rather undesirable to large swathes of the electorate of the current government majority.