AI also reads POS and receipts to combat tax evasion and fraud: analysis of integrated databases to target controls, but it will be man who decides

tl;dr:

the 200 databases will be read and integrated by algorithms to find tax evaders and increase tax compliance.

I think the article is behind paywall, copy-paste:

Giorgetti signs the guidance document on the 2026-2028 objectives for Agencies, Gdf and Finance department. Analysis of databases integrated with artificial intelligence to target controls, but man will decide

The fight against tax evasion will aim to increasingly divide honest taxpayers to be accompanied in compliance and those at risk on whom controls should be concentrated. Greater help will come from artificial intelligence, which will be used to read in an integrated way the boundless amount of information present in the 200 tax databases, which are gradually more integrated with each other, and define the risk profiles to guide the controls: which however will continue to be decided by officials and managers of the financial administration, based on the indispensable principle of “non-exclusivity of the algorithmic decision”.

Artificial intelligence will enter into the preparation of formal checks on declarations, and will also lend a hand to speed up refund procedures.

This is explained by the 82 pages of the 2026/28 fiscal policy document signed on 25 February by the Minister of Economy Giancarlo Giorgetti. The program starts from the claim that the fight against undeclared work is not at year zero. Between 2019 and 2022, the propensity to evade taxes (net of IMU and excise duties) reduced by 11.7%, therefore exceeding the target of a 10% cut set by the Pnrr. And on this path the Government has moved forward with a wide range of measures, from interventions against VAT violations and undue compensation to the extension of the information assets of the Revenue-Collection Agency, up to the obligation to align POS and receipt data.

The risk profiles

One of the lines of action traced by the new plan starts right from here. Because “the obligation to connect the information deriving from electronic payments and the registration of fees”, we read in the document, offers one of the key chapters of the Sif, the Tax Information System, on which “the assessment of the risk of non-compliance and the monitoring of subjects at high fiscal risk will be based”.

The definition of risk profiles, which will be conducted jointly by Revenue, Customs and GdF to avoid overlaps, will be based on two sets of factors. On a general level, “disaggregation on a provincial basis” and “articulation according to the different economic sectors” will have an impact. But in this grid a series of individual behaviors will then count, such as the habit “of submitting tax returns promptly, paying taxes punctually and not transmitting unfaithful, incorrect or incomplete declarations”.

I repeat offenders

An average higher risk is attributed to those who have already encountered an inspection that has revealed irregularities. Consequently, “the Agency will monitor the behavior of taxpayers who have undergone a tax audit to verify over time their degree of propensity to comply and, more generally, to evaluate the level of tax fidelity”. On this basis, the policy document calls for “a greater frequency of substantial checks”, here too “using electronic invoicing data, current account registers, electronic payments and digital receipts in a predictive key”. And the mantra of prevention also guides the project to strengthen, again through risk analysis, preventive controls on credit transfers and compensation F24s.

The drive for compliance

Technology and AI will also aim to strengthen the drive for compliance, a branch on which the Government has bet as a lever to change the relationship between tax authorities and taxpayers. In fact, on the agenda is the improvement of the “selection algorithms” of the recipients of compliance letters, to reduce “false positives”, and the analyzes on the effectiveness of these initiatives, in terms of revenue recovered, will be developed with “data analysis tools”. Moreover, technology will guide the entire cost/benefit analysis of anti-evasion initiatives, through new “aggregate indicators of increase in tax revenue” to “measure the increase in revenue deriving from prevention and enforcement activities compared to 2024”.

The list of homework also affects the Customs agency, called to combat the “most dangerous phenomena” identified in particular in “total duty evasion, under-invoicing and import VAT fraud”, also implemented through e-commerce and the use of split shipments (mini parcels).