Maria Grazia Ortoleva
Associate Professor of Tax Law
Law Departement
University of Verona (Italy)
Revista Técnica Tributaria, Nº 139, Sección Estudios, Cuarto trimestre de 2022, AEDAF
Inteligencia artificial: una herramienta «peligrosa» en manos de la administración fiscal italiana para luchar contra la explotación abusiva de las deducciones fiscales
El uso de la IA está provocando, también en Italia, una metamorfosis de la relación entre el contribuyente y la Administración tributaria tanto en la fase de aplicación de los impuestos (por ejemplo, los borradores del impuesto sobre la renta y del IVA) como en los procesos de verificación y control.
Este cambio, si no se «gobierna» adecuadamente, podría sin embargo tener un alto coste en términos de vulneración de algunos derechos fundamentales de los contribuyentes, sobre todo cuando se «confía» a sistemas de inteligencia artificial la adopción de decisiones que tienen efectos inmediatos sobre la situación jurídica de los destinatarios.
Esta es la característica que distingue a la medida introducida por el legislador italiano para atajar el fraude perpetrado a través de la transferencia de créditos fiscales por intervenciones destinadas a mejorar la eficiencia energética de los edificios y que es objeto de investigación en este trabajo.
Inteligencia Artificial, vulneración de derechos, nuevas tecnologías, algoritmos, fraude, créditos fiscales
The use of AI is causing, including Italy, a metamorphosis of the relationship between the taxpayer and the tax administration both in the tax implementation phase (for example the pre-filled income tax and VAT returns) and the verification and control processes.
This change, if it is not properly governed, could however have a high cost in terms of infringement of some fundamental rights of taxpayers, above all when the adoption of decisions that have immediate effects on the legal status of the recipients is entrusted to artificial intelligence systems.
This is the characteristic that distinguishes the measure introduced by the Italian legislator to tackle fraud perpetrated through the transfer of tax credits for interventions aimed at improving the energy efficiency of buildings and which is the subject of investigation in this paper.
Artificial Intelligence, infringement of rights, new technologies, algorithms, fraud, tax credits
Fecha de recepción: 14-12-2022/Fecha de aceptación: 19-12-2022/Fecha de revisión: 2-01-2023
Cómo referenciar: Ortoleva, M.G. (2022). Artificial intelligence: a «dangerous» tool in the hands of the Italian tax administration for fighting the abusive exploitation of tax relief. Revista Técnica Tributaria (139), 175-196
1. Introduction
In recent years, tax administrations around the world have increased their investment in IT tools. The goal is to make use of «new» technologies to «administer» taxes, support taxpayers and improve their level of «compliance». Italy is no different. The tax administration’s use of technologies and, in particular, of big data, algorithms and artificial intelligence systems has grown in the last decade and is destined to grow further.
There are many reasons behind this choice.
In summary, it can be said that «algorithms» are considered not only as being more efficient than human analytical skills, but also that they have an aura of neutrality. (1)
Algorithms are, in fact, capable of analysing very large quantities of data, even if they are heterogeneous from a qualitative point of view, identifying valuable correlations, relationships and inferences.
Furthermore, to the extent that they arrive at «decisions» which are the result of purely rational calculations based on data, they should make it possible to correct the distortions and imperfections that typically characterise cognitive processes and the choices made by human beings. Finally, state investments in artificial intelligence systems have proved to be very profitable. (2)
However, the enormous potential inherent in the application of «intelligent systems» also entails risks. (3)
In fact, experience has already highlighted certain dangers that can arise from the use of algorithms. The results of artificial intelligence systems can be erroneous and misleading, for example, due to incorrect data entry or bias in the construction of mathematical models or errors within their operation. (4)
It is therefore necessary to evaluate the impact that these tools have on people and their fundamental freedoms and carefully weigh up their use. (5)
The objective of this paper is to investigate the implications arising from the Italian tax administration’s use of artificial intelligence to fight frauds connected to the use of tax credits and, in particular, to that perpetrated through the transfer of tax credits due for the implementation of specific interventions aimed at improving the energy efficiency of buildings (known as the «Superbonus»). (6)
In particular, we intend to verify whether the measures adopted by the Italian legislator make it possible to counter the aforementioned fraud without compromising and/or limiting the fundamental rights of taxpayers.
To this end, the first paragraph will briefly describe the «favourable» discipline introduced in Italy to facilitate investments aimed at obtaining energy efficiency. The second paragraph will examine the specific measures to combat fraud introduced by the legislator with particular reference to those that resort to the use of AI. Finally, starting from the precedents of internal administrative case law and in the light of the principles contained in the GDPR as well as the general principles of the legal system, it will be verified whether the solution prepared leads to a satisfactory result in terms of balance between the apparently opposing needs for the protection of the state's fiscal interest and the protection of the taxpayer's rights.
2. The «favourable» discipline introduced in Italy to improve the degree of energy efficiency
Since 2020, the Italian legislator resorted to tax leverage to improve the degree of energy efficiency of the Italian real estate assets and to foster economic recovery.
In particular, the tool used was that of granting to natural persons (acting outside the exercise of business, art and profession) (7) a tax credit equal to 110% (8) of the expenses incurred to carry out specific interventions aimed at improving the energy efficiency of buildings (9) (known as the «Superbonus»).
This credit can be used by the taxpayer in the tax return to reduce the tax liabilities deriving from the same tax return. In particular, the deduction of the tax credit must be divided into four equal annual instalments and can be used within the limits of the annual tax liabilities deriving from the tax return.
The deduction is not the only way to benefit from the tax relief.
In fact, it is stated that, instead of direct use in the tax return, the beneficiaries (credit holders) can opt alternatively for:
Both the suppliers (who have applied the discount on the invoice) and the transferees can, in turn, transfer the «purchased» tax credit to other parties or use it to offset the taxes they owe.
The method of use of the credit chosen by the taxpayer must be communicated electronically to the Revenue Agency, using a specific form (11) (comunicazione web cessione crediti/sconti or credit transfer/discount web communication). (12)
A similar obligation of communication is required for suppliers and assignees who decide to transfer the «purchased» credit to third parties». (13)
It is easy to understand that the provision of this particular tax relief, together with the possibility (also for transferees and suppliers) of transferring the tax credit, has exposed the system to many fraudulent activities.
The Bank of Italy, at the end of 2020, highlighted the existence of fraud and money laundering risks due to: the possible fictitious nature of the receivables, the presence of assignees who pay for the assignment with capital of possible illicit origin, and the «abusive» carrying out of financial activity by persons without the required authorisation.
The existence of fraudulent behaviour has been confirmed by the investigative and analytical activities of the Guardia di Finanza (Italian Finance Police). (14) This included, among other things, «chain» transfers of tax credits involving companies with the same registered office and/or with the same legal representatives, whose partners or directors are destitute, untraceable and/or have a criminal record (15) ; fractional assignments of the same credit by the same subject to conceal their probably fictitious nature; the non-attribution to the original beneficiary (credit holders) (transferor) of the property on which the subsidised interventions would have been carried out; the incompatibility of the building works with the dimension of the company that would have carried them out and that acquired the credit with the «invoice discount».
In order to stem these activities, specific measures aimed at strengthening preventive control against fraud committed in this area have been adopted in the years following the entry into force of the tax relief (16) . The legislator has intervened on several fronts and in particular: by introducing limits on the assignment of credits; anticipating the intervention, before each use of the credit, of professionals who certify, on the basis of the documentation produced by the taxpayer pertaining to the intervention, the existence of the conditions that give the right to the tax deduction (known as the stamp of approval) (17) ; by strengthening the tools and control powers of both the finance police and the tax administration through the use of artificial intelligence (understood in a broad sense).
In particular, under the latter profile, the finance police have been given two new applications, called «P.Ris.MA.» and «Moni.C.» in order to make verification and control processes more timely, including preventive audit, aimed at combatting fraud perpetrated in this area, and a new and specific control procedure was introduced, intended for the transfer of tax credits or invoice discounts.
This procedure, making use of the calculation capacity of the AI, allows the Revenue Agency to identify «credit transfer/discount communication» that present risk profiles and to temporarily suspend them, for the purpose of carrying out a further «preventive» check (i.e. before the time from which the tax credit can be used by the transferee or supplier who applied the discount on the invoice).
The following investigation will focus precisely on this last measure, not only on account of the novelty but above all of its effects. We refer to the fact that the use of AI, in the case, does not only represent a tool for risk assessments and detecting fraud the selection of taxpayers and operations (assignments and discounts) to be subjected to timely audits but it can also lead to automated (that is, without human intervention) adoption of an administrative act able to inhibit, albeit temporarily, the «tax» effects of the assignment of credits or the «discount on the invoice», i.e. the usability of the credit by the transferee.
This possible outcome of an automated procedure raises the possibility that, in this context, certain taxpayers' rights may be «sacrificed» due to the need to combat fraud (18) . For this reason, therefore, it is necessary to investigate this aspect and to verify whether this is founded.
To this end, as a preliminary point, it seems appropriate to briefly consider the reference discipline of the new control procedure.
3. Preventive audit of «credit transfer/discount communication»
It has already been said that this measure is one of those introduced with the Antifrode Decree. In particular, it is governed by art. 122 bis of Law Decree no. 34/2020. This rule gives the Revenue Agency the power to suspend the effects of «credit transfer/discount communications» (even subsequent to the first one) which present risk profiles within five working days of sending the communication (19) and for a period not exceeding 30 days.
The identification of communications at risk (i.e. profiling) takes place on the basis of criteria that are identified, albeit in a very general way, by the legislator. In particular, it is provided that for this purpose, in relation to the different types of credits, the following elements are taken into consideration: a. consistency and regularity of the data indicated in the communications with that present in the Tax Registry or in any case held by the tax administration; b. data pertaining to the credits being transferred and to the subjects involved in the transactions (related to these credits), on the basis of the information present in the Tax Registry or in any case in the possession of the tax administration; c. similar transfers carried out previously by the subjects indicated in the communications.
If the audit carried out on the basis of these criteria shows that the operation is at risk, the effects of the communication are automatically suspended for a maximum period of thirty days. This suspension is communicated via the telematic services of the Revenue Agency to the person who sent the communication. (20)
From that moment, the thirty-day period within which the tax administration must carry out a more precise audit begins, intended to ascertain (presumably, with human intervention), the «substantiation» of the elements of risk.
In reality, nothing is stated by law regarding the methods of this second audit, nor is it specified by the implementing provision. (21)
The second paragraph of Article 122 bis regulates only the possible effects. In particular, it states that if, as a result of the audit carried out, the elements that led to the suspension are confirmed («gli elementi che hanno determinato la sospensione sono confermati»), the communication is considered not to have been made («la comunicazione si considera non effettuata»).
Paragraph 2 continues by stating that: if instead, the risks are not confirmed, the communication produces the effects stated by the reference legislation. The same happens if, within the suspension period, the subject does not receive notice of the voiding of the communication.
On this point, the implementing provision follows the provisions of art. 122 bis and adds that, in the event that the audits carried out confirm the legitimacy of the suspension, the voiding of the effects of the communication, with the relative reason, is communicated (via the telematic services of the Revenue Agency) to the person who has sent the communication.
Furthermore, in this case, by express provision of the law, the tax administration proceeds, in the terms of the law («nei termini di legge»), to audit the related credits, i.e. first of all on the right to tax relief. (22)
It should be noted that, making use of the powers provided for by articles 31 et seq. of the Presidential Decree no. 600 of 1973, the tax administration can always ascertain the existence of the conditions for the entitlement of the tax deduction (for the person who presented the communication) and the «regularity» of the transfer transactions (and discount). (23) In particular, given the wording of the provision referred to in paragraph 3 (and in particular the use of the term «Fermi restando gli ordinari poteri di controllo»), the right of the taxpayer to the tax relief remains subject to the ordinary substantial audits both when the administrative measure of suspension of the «credit transfer/discount communication» is subsequently «annulled» by the tax authority, and when its effects are void due to the expiry of the 30-day term. In such cases, therefore, no legitimate expectation as to the right to tax relief (and consequently on the regularity of the assignment or discount) can be formed either by the direct beneficiary or by the transferee/supplier. (24)
Furthermore, according to the tax administrations, the existence of this preventive control does not exempt the subjects involved in the transfers — primarily, the transferees and suppliers — from using the ordinary diligence required to avoid participating in fraudulent conduct. (25)
This, in summary, is the reference framework of the preventive audit of the «credit transfer/discount communication». It clearly shows the elements that distinguish the measure in question.
Indeed, with art. 122 bis, the legislator seems to have introduced a sort of «automated» audit of «credit transfer/discount communication» into the legal system. This control, carried out through the use of algorithms and the analysis of big data, may culminate, firstly, in the adoption of an algorithmic act that authoritatively determines the temporary suspension of the effects of the «communication» (i.e. the choice of the taxpayer to transfer the credit to third parties or to the supplier) and, then, in the voiding of the «communication».
The control procedure appears, in fact, divided into two sub-procedures (phases), the first of which is entirely automated.
In particular, in the first phase, the tax administration makes use of the computing power of AI to analyse «communications» in a very short period of time, to intercept those expressions of evident and «significant» illegal conduct and suspend the effects of the latter before the credit can be used by the supplier/transferee.
In the second phase, however, the control concerns only «suspended» communications. It is aimed at verifying the elements of risk and can lead to the voiding of the communication.
The law does not state, whether this procedure is also «automated» or whether «human» intervention is required, as it should be if the intent is actually to carry out a verification, albeit documentary, of whether the risk elements on the basis of which the «communication» was suspended are in fact true. (26)
In summary, this is a control procedure which, in addition to being automated, is likely to make choices made by taxpayers on how to use the Superbonus ineffective, without them having complete knowledge of the reasons for the administrative activity.
4. Critical profiles
Reading the regulatory framework governing the measure in question seems, in essence, to confirm the suspicion that the exercise of the control stated may infringe certain fundamental rights of taxpayers.
In that sense, there are several critical issues that, on the whole, emerge from the discipline. These start with the opacity of the criteria that can determine, firstly, the automatic and temporary suspension of the effects of the «credit transfer/discount communication», then, the voiding of the same (communication), up to the failure to provide the reason of the administrative acts (27) and cross-examination with the subjects involved in the operation in various capacities.
For the purposes of the investigation, it may be useful to compare this measure with another «form» of automated audit existing in the legal system, concerning income tax returns pursuant to art. 36 bis of Presidential Decree no. 600 of 1973 (known as Liquidazione of income tax returns). (28)
The preventive control pursuant to art. 122 bis presents, in fact, characteristics similar to those of the automated control of income tax returns but, at the same time, differs from it in several respects.
In particular, similarly to the Liquidazione of income tax returns, the preventive audit on communications relates to all the «credit transfer/discount communications» and appears to be an «internal» control, so to speak. From the legislative text it appears that it is carried out, first of all, on the basis of the comparison between the consistency and regularity of the data indicated in the communications and the data in possession of the tax administration possibly present in the Tax Registry. It also takes into account further information, always in the possession of the tax administration, relating to credits being transferred and to the subjects involved in the transactions or to similar assignments carried out previously by the subjects indicated in the communications.
Furthermore, similarly to the Liquidazione of income tax returns, it also does not preclude, as mentioned, the exercise by the tax administrations of the ordinary substantive control powers in order to ascertain the right to tax relief and the «regularity» of the assignments.
As for the differences, first of all, the audit of the «credit transfer/discount communication», unlike the Liquidazione of income tax returns, in fact, seems to lead to, and presuppose, a profiling of the transferors or transferees. (29)
But then and above all, it differs clearly from the automated one of income tax returns as regards the effects in the event that «anomalies» are found.
In this case, in fact, in the Liquidazione the tax administration sends the taxpayer an administrative act, know as the, «avviso bonario», containing an indication of the irregularities found and the reasons for any recovery of the tax, in order to allow the recipient to regularise their position, including in contradictory procedure with the office (30) .
The control in question, instead, leads to a real «algorithmic» administrative measure of a precautionary nature. This measure on one hand, is the result solely of the use of AI systems, and on the other, authoritatively determines the temporary suspension of the effects of the «credit transfer/discount communication» (31) , i.e. the usability of the credit by the transferee, pending the outcome of the subsequent audit and for a predetermined time. Despite its administrative measure nature, art. 122 bis does not provide that the suspension must be reasoned (not even by a reference to the elements which, among those identified by the legislator, have led in the specific case to the qualification of the transfer/discount as with a high-risk profile of fraud and to the consequent suspension), nor does it recognise the right to provide clarifications to the taxpayer/person who submitted the communication. Given the wording of art. 122 bis, it seems that the tax administration can «autonomously» proceed with the verification of the legitimacy of the suspension and with the eventual voiding of the «credit transfer/discount communication».
In this regard, as anticipated, it is not clear whether the possible voiding of the communication is the result of a further «automated» procedure, nor in any case, is it clear on the basis of which facts/documents/criteria the tax administration can confirm or otherwise of the elements that led to the suspension.
In conclusion, the foregoing analysis confirms that the regulation of the preventive control of communications lacks certain guarantees (e.g. reason of the interim administrative measure (avviso bonario) (32) , intra-procedural right to be heard (33) ) present, instead, in art. 36 bis. These guarantee the impartiality, fairness and transparency of the administrative action or, in summary, the respect of the right to good administration provided for by art. 41 of the Charter of Fundamental Rights of the European Union (34) .
The absence in art. 122 bis of provision of adequate guarantees (which in the case would have been even more necessary due to the existence of the automated decision) means that not insignificant abridgement of the right of taxpayers to have an administrative activity that complies with the principles of good administration can result from the «automated» control.
These limitations do not seem insignificant because they concern the essential rights of democratic societies. Moreover, at first sight, they seem difficult to justify, both by reason of the «neutrality» of the algorithmic decision and regarding the need to combat fraud or to safeguard tax revenue.
Indeed, none of these «exceptions» seem in themselves able to make plausible or acceptable the aforementioned abridgement of the rights of the subjects affected by the measures adopted in the context of the preventive control pursuant to art. 122 bis.
The delicacy of the matter requires further study, which is essential to attempt to formulate an assessment, even if it is not definitive.
5. The administrative case law on the «algorithmic administrative act»
As for the «neutrality» of algorithms, the question concerning the implications deriving from their use in administrative procedures and, in particular, that inherent the (un)lawfulness of the abridgment of citizens' rights caused by algorithmic decisions is by no means new in the legal field.
In Italy administrative jurisprudence has dealt with it, although not in tax matters, when it was called upon to judge the legitimacy of the «algorithmic administrative act». (35)
For the purposes of this investigation, it therefore seems appropriate to reference the criteria elaborated by this case law.
The Italian Council of State has ruled on the issue on several occasions that all originate from the same story: the use by the Italian Ministry of Education of an algorithmic system for defining the assignments of secondary school teaching staff.
In these cases, the judges, first of all, recognised the need to take advantage of the significant potential of the digital revolution», emphasising that the use of algorithms in administrative procedures is certainly admissible (36) , and should also be «encouraged», since it is able to ensure that the principles of impartiality and sound administration are respected pursuant to art. 97 of the Constitution. (37)
At the same time, however, the judges of Palazzo Spada subordinated the usability of algorithmic systems with respect to specific conditions (which represent elements of minimum guarantee for any situation of use of algorithms in public decision-making (38) ). In particular, these were the «full knowledge of the IT module and the criteria used» (39) and the «imputability of the decision to the body holding the power, which must be able to carry out the necessary verification of the logic and legitimacy of the choice and of the outcomes entrusted to the algorithm». (40)
The judges of Palazzo Spada started from the assumption that the algorithm should be understood as a «procedural and preliminary tool, subject to the typical checks of any administrative proceeding», and came to the conclusion that its use remains subject to the same guarantees as administrative activity. Among these the Judges pointed out, the knowability and comprehension of the technical regulation in all its aspects, «from its authors to the procedure used for its elaboration, to the decision mechanism, including the priorities assigned in the evaluation and decision-making procedure and the data selected as relevant». (41)
According to this orientation, therefore, for the purpose of the legitimacy of the algorithmic administrative act, the mere knowledge of the procedure used for its elaboration and of the decision mechanism is not sufficient, but the effective comprehensibility of the algorithm, as well as its pertinence to the legal rule, must be guaranteed.
The judges stated that the mathematical formula must be understandable to everyone, starting from the body holding administrative power (to which the administrative decision must be attributed), and must be accompanied by explanations that demonstrate its pertinence to the «legal rule» it underlies.
These conditions are considered essential. Indeed, compliance with them ensures that:
The case law of the Council of State confirms, in summary, that, although the algorithms are characterised by an aura of neutrality, their use in making public decisions cannot in any way abridge the right to impartiality, fairness and transparency of administrative action. On the contrary, it reinforces the need for transparency of the administrative activity, which can be said to be guaranteed if all the subjects involved in the proceeding are able to know and understand the technical rule underlying the legal one (43) .
On the basis of this, therefore, it seems to be corroborated that the use of algorithmic systems by the tax administration in tax control and assessment procedures and in the fight against fraud is certainly possible and desirable (44) , but it cannot be detrimental to the aforementioned minimum guarantees of administrative activity.
So, turning our attention to the preventive audit of «credit transfer/discount communication», compliance with these guarantees would presuppose the knowability, comprehensibility and pertinence both of the criteria and of the decision mechanism on the basis of which the communications are suspended and of those that can lead to the voiding of the communication.
The analysis of the discipline dictated by art. 122 bis demonstrates, however, that in this case, the need for transparency is not guaranteed at all.
Paraphrasing the words of the Council of State, the knowledge of the IT module used and of the criteria is far from complete in both sub-procedures in which the audit is divided.
Article 122 bis refers to notions (those of «regularity» and «consistency») and facts («other analogous transfers») that are only apparently clear. Indeed, what is clear and certain in vitro is not always so in vivo.
Regularity and consistency, for example, are notions which can have various meanings and can be satisfied to different degrees, so it is not at all easy to establish when there is an effective and full response.
It has already been noted that the selection criteria are indicated in a very general way.
It must now be added that the vagueness of the risk elements is demonstrated by the regulatory formulation itself. Paragraph 5 refers their specification to a provision of the Director of the Italian Revenue Agency. The provision was issued but it is not helpful as it exactly follows the text of art. 122bis. No unequivocal cases were indicated which determine the possibility of suspension; the meaning of the terms employed is not explained; and the significance and pertinence of the various risk elements were not specified.
The process through which the adoption of the suspension measure is reached on the basis of these elements remains completely obscure: do they all have the same weight or are some overriding? This is one of the many questions that arise.
The opacity of the selection criteria and the lack of transparency on the decision mechanism underlying them, together with the tendency of people to automatically and uncritically rely on AI and its results, also raise doubts about the «effectiveness» of the verification that the tax administration is required to carry out following the automated measure of suspension of communication.
How can the tax administration, in fact, verify the logic and legitimacy of the choices made by the algorithm and possibly confirm the decision, voiding the «credit transfer/discount communication», if it does not know and understand the algorithmic rule in all its aspects? (45) If then the communication of option or transfer is voided, how can the tax administration specify reasons for its decision?
Furthermore, the question on the effectiveness of «human oversight» is present in the proposal for a regulation on artificial intelligence (Artificial Intelligence Act). (46) In this proposal the Commission recognises that the mere attribution of a right to a decision resulting from human verification is destined to remain without substance if not assisted by a series of measures which make its applicability concretely possible (47) .
The original lack of transparency and comprehensibility runs the risk of depriving of meaning both any human intervention and the reason for the decision to void the communication. In summary, the analysis carried out so far shows that there is a risk that the preventive audit is exercised by the Revenue Agency in an «arbitrary» way, i.e. not compliant with the principles also expressed by the Council of State.
The fact that the output cannot be influenced by the executor (since the same results are always obtained with the same data) does not guarantee the impartiality of the administrative activity. (48)
The tax administration has the obligation to verify that the criteria, conditions and results of the automated procedure comply with the rules and purposes established by law (49) . In the same way, the judge, if called to rule, must check the measures of the tax administration like any other administrative measure and, in order to exercise this function, must be able to know and evaluate the logical legal process underlying the act. In fact, the principle according to which there is no presumption of the legitimacy of tax administrative measures is undisputed.
On the basis of the above arguments, therefore, it seems that the exercise of the audit pursuant to Article 122 bis can cause abridgement to the principle of transparency of the administrative action, and that such abridgements cannot be considered «acceptable» either by reason of the neutral nature of the algorithms or the uniqueness of tax matters and the need to safeguard public finances.
6. The relevance of the principle of legal certainty
In the latter respect, from a general point of view, it should be added that the pursuit of tax interest may admit exceptions to common, civil, administrative, criminal law and may allow the legislator to limit the rights of individuals, but in compliance with the principles of the rule of law, including that of legal certainty. (50)
In the tax field, unlike other sectors of the legal system, for example the civil (51) one, the instances connected to the idea of legal certainty have far from been obliterated. They have always been present, not only in the doctrinal debate (52) , but also especially in community case law. (53)
According to the EU Court of Justice «Individuals must be able to ascertain unequivocally what their rights and obligations are and take steps accordingly» (54) ; and that necessity imposes itself with particular rigour when it comes to legislation regarding financial burdens. (55) In this context, therefore, the objective predictability of the application of the rules must be guaranteed.
The Court also attributes to the objective dimension of legal certainty the value of a «general principle» or «founding principle» of the Union, inherent in the rule of law whose principles must be followed by the European Union and the Member States after the Treaties of Amsterdam and Lisbon. (56) Certainty is, therefore, an objective for which every State must necessarily strive, first of all, to guarantee the freedom of individuals and the equality of citizens under the law.
Specifically, according to the Court, this principle requires that rules of law are known, clear, precise, stable, certain and predictable. (57) Regulatory texts must be drafted with rigour and precision, in terms that are not vague or ambiguous, nor open to interpretation, so as to allow the interpreter-applicator to derive non-contradictory rules. (58)
Going back to the question of the limits within which the rights of taxpayers subject to automated control procedures can be constrained, one can derive the following from this assumption. The tax provisions which, to combat fraud and safeguard public finances, envisage the use of algorithms in the tax field (and which therefore, among other things, «authorise» the automated processing of personal data and the adoption of automated decisions) must not only comply with the principles established by GDPR (59) but, first of all, be formulated in such a way as to make their application predictable.
The latter precept was recently expressly cited by the Slovak Constitutional Court which, in declaring the incompatibility with the Constitution of the legislation that allowed the use of AI systems without effective supervision, stated that « The law restricting fundamental rights must be specific enough to make its application predictable» (60) . According to the Court, not only «The application of technological progress in public administration cannot result in an impersonal state whose decisions are inexplicable, unexplained and at the same time no one is responsible for them» (61) , but, even before that, «the legislation must provide the citizen with a sufficiently clear indication of the circumstances and conditions under which the public authority is empowered to interfere with their rights». (62)
Even from this point of view, the measure adopted by the Italian legislator is not immune to criticism.
The «gaps» concerning the criteria on the basis of which the algorithm analyses the risk of fraud, not filled by the provision of the Director of the Revenue Agency, do not allow to objectively predict the conditions under which the tax effects of the «communications» can be suspended. In other words, an unjustifiable margin of uncertainty remains, regarding the reasons that can stop the tax credit from being used by the transferee.
As stated by authoritative doctrine, it is not claimed that tax laws are written in such a way that their interpretation is certain, i.e. automatic (63) , but that their formulation is clear enough to allow everyone to understand their effects.
7. Conclusion
There is no doubt that new technologies can offer a valuable contribution to the fight against tax evasion and fraud and, ultimately, to the achievement of the tax fairness which in the Italian legal system underlies the very idea of progressive taxation enshrined in the Constitution.
At the same time, it is equally undoubted that the «innovation» process must be put on the right track so that it generates effective benefits for the community.
In the tax field, the use of AI is causing a metamorphosis in the relationship between the taxpayer and the tax administration, both in the tax implementation phase (for example the pre-compiled tax returns and VAT) and in the audit phase.
In Italy, this change — as demonstrated by the investigation carried out here — has not, however, so far been accompanied by the introduction of adequate guarantees for the new context and not even by an adaptation of the existing (64) and hard-won ones. (65)
Even the most recent interventions on the subject of processing, through risk analysis systems (known as Ve.r.a. anti-evasion algorithm), of the personal data of taxpayers contained in the archive of financial reports take in the same direction, essentially limiting themselves to reducing the extent of the rights provided for on the protection of personal data by Regulation (EU) 2016/679 of 27 April 2016 («GDPR»). (66)
This tendency can only generate concern.
Experience, including national, shows that the use of AI systems by public authorities can have a negative impact on the fundamental rights of the recipients.
It is therefore necessary to look for solutions that maintain a balance between the protection of the state interest and the rights of taxpayers, and this task falls primarily to the legislator. (67)
It can be said that the real challenge is «to govern with artificial intelligence by governing artificial intelligence».
Technocratic shifts must be avoided, guaranteeing the development of AI systems designed to be used without compromising the transparency of the administrative action, i.e. systems that are transparent, explainable and documented.
At European level, an attempt is underway to create a «framework for trustworthy AI».
The proposal for a regulation on artificial intelligence (Artificial Intelligence Act), in aiming to create an «ecosystem of trust» that puts people at the centre, seems to be moving in the right direction. However, it must be pointed out that even in this context it is stated a priori that «AI systems specifically intended to be used for administrative proceedings by tax and customs authorities should not be considered high-risk AI systems used by law enforcement authorities for the purposes of prevention, detection, investigation and prosecution of criminal offences». (68)
See Cons. State, Section VI, Judgment, 13 December 2019, no. 8472, in Nuova Giur. Civ., 2020, 4, 809.
In the United Kingdom, for example, the artificial intelligence system, known as Connect, used by the HMRC since 2010, allegedly cost £100 million and helped recover more than £ 3 billion in taxes. See Sanghrajka, J. (14 July 2020), HMRC’s Connect Computer and Investigations, Taxation 4752, available at https://www.taxation.co.uk/articles/hmrc-s-connectcomputer-and-investigations (accessed 10 Oct. 2022).
Regarding this see de la Feria, R. — Grau Ruiz, M.A. (2022), The Robotisation of Tax Administration. En M. A. Grazu Ruiz (coord.) Interactive Robotics: Legal, Ethical, Social and Economic Aspects, 115-123, Springer; Pitruzzella, G. (2022), Dati fiscali e diritti fondamentali. Diritto e pratica tributaria internazionale (2), 666-677; Messina, S.M. — Ortoleva, M.G. (2022), Artificial Intelligence and Robots: The Role of Tax Legislator, A Conundrum to Solve. En Interactive Robotics: Legal, Ethical, Social and Economic Aspects, cit., 135-142; Grau Ruiz, M.A. (2022), Fiscal transformations due to ai and robotization: where do recent changes in tax administrations, procedures and legal systems lead us? Northwestern journal of technology and intellectual property (19), 4, 325-363; Kuźniacki, B. — Almada, M. — Tyliński, K. — Górski, Ł. — Winogradska, B. — Zeldenrust, R. (2022), Towards eXplainable Artificial Intelligence (XAI) in Tax Law: The Need for a Minimum Legal Standard. World Tax Journal, Vol. 14, No. 4, 1-28; Pinto Nogueira, J.F. (June 2, 2022). Tax Administration and Technology: from Enhanced to No-Cooperation? Digital Transformation of Tax Administrations, available at https://ssrn.com/abstract=4125999; Dorigo, S. (2021), The 'Algorithmic Revolution': Fair Taxation, Social Pact and Global Governance. En M. Belov (coord.), The IT Revolution and its Impact on State, Constitutionalism and Public Law, 161-179. Santoro, A. (2020), Lotta all’evasione fiscale: i big data e la nuova frontiera della prevenzione. Aggiornamenti Sociali, 190-199.
To mention just two examples: the judgment in Loomis v. Wisconsin (US: Wis. SC, 13 July 2016, Loomis v. Wisconsin, 881 N.W.2d 749) and the SyRI case (Rechtbank Den Haag, 5 Feb. 2020, NJCM et al. v. Netherlands, NL:RBDHA:2020:1878, in https://uitspraken.rechtspraak.nl/#!/details?id=ECLI:NL:RBDHA:2020:1878.
Pollicino, O. (2021), Getting the Future Right — Artificial Intelligence and Fundamental Rights». A view from the European Union Agency for Fundamental Rights. BioLaw Journal (1), 7-11, stresses that «The first aspect that any regulations adopted in the area of AI must take account of is their impact on fundamental rights as enshrined in the Charter and the EU Treaties». Frosini, T.E. (2020), Il costituzionalismo nella società tecnologica. Dir.Inf., 465-484.
See art. 119 of Law Decree no. 34 dated 19 May 2020 (converted by Law no. 77 dated 17 July 2020) kwon as the Rilancio Decree.
Also, other subjects can benefit from the credit, such as condominiums, non-profit organisations (non-profit organisations of social utility), voluntary organisations and social promotion associations registered in the appropriate registers, provided that these entities meet certain requirements and within specific spending limits. Persons subject to corporate income tax (IRES) are therefore excluded from the Superbonus (e.g., commercial companies). See art. 121 of Law Decree no. 34/2020. Regarding this see Del Federico, L. (2022), Profili attuali in tema di crediti d’imposta: poliformismo, funzione sovvenzionale, tutele e finanziarizzazione. Rivista di diritto tributario (3), I, 201.
The amount of the tax credit was subsequently decreased according to the beneficiaries and the tax period in which the expenses were incurred. See 2022 Finance Act.
In particular, it must be thermal insulation interventions on the enclosures; replacement of winter air conditioning systems in the common areas; replacement of winter air conditioning systems on single-family buildings or on functionally independent multi-family buildings (so-called «leading» or «main» interventions). The beneficiaries can benefit from the tax deduction also for the expenses incurred for all the energy efficiency measures referred to in the so-called Ecobonus and for the installation of infrastructure for the recharging of electric vehicles and solar photovoltaic systems («minor» interventions), provided they are carried out together with at least one of the «leading» interventions. See paragraphs from 1 to 4 of Article 119 of the Law Decree no. 34/2020.
See art. 121, co. 7, of Law Decree no. 34/2020. Within 5 days of sending the communication, the agency's computer system issues a receipt of acceptance or rejection of the request. For expenses incurred in 2021, the deadline within which the option must be communicated is 29 April 2022. See art. 10 quater Law Decree 27 January 2022, no. 4 (know as the Aiuti ter Decree). December 1, 2021, a provision from the Director of the Italian Revenue Agency (amended by subsequent provisions) approved the details for exercising options regarding eligible deductions and, specifically, the model for reporting the option.
They must, first of all, accept the transactions that have taken place against them and then communicate the method of using the credit they have chosen to the tax administration. These communications must be made through a specific electronic procedure, by accessing an application entitled Piattaforma Cessione crediti.
Suffice it to say that the financial police with a single operation discovered a fraud of a total amount of € 103,067,709.00.
Generally, these are companies established in a short period of time or which have resumed operations after a period of inactivity.
The stamp of approval can only be issued by the subjects indicated in article 3, paragraph 3, letters a) and b) of Presidential Decree no. 322 of 22 July 1998. This includes those registered in the registers of chartered accountants and accounting experts and labour consultants.. In the case of the Superbonus, the audit carried out by the professionals are a mere formal documentary check. As a result of the changes introduced with the 2022 Financial Act, the taxpayer must request the stamp of approval both to use the deduction in their tax return and to opt for the transfer or «invoice discount».
See Scarcella, L. (2019), Tax compliance and privacy rights in profiling and automated decision making. Internet Policy Review (8), 4, 1-19, according to which «safeguards to taxpayers’ privacy rights need to be in place»
Pursuant to art. 121, para 7, of Law Decree no. 34/2020 the exercise of options must be carried out electronically.
See Provision of the Director of the Italian Revenue Agency, 1 December 2021, point 3.1.
See art. 122 bis, para. 3, of Law Decree no. 34/2020. See also Explanatory Report on the Antifrode Decree, where it is clarified that the provisions of art. 122 bis shall perform preventive control functions, without prejudice to the powers of control over the merits of the tax administration in relation to credit transfer transactions that took place both before and after the entry into force of the law.
According to the Revenue Agency the non-selection of the specification among those subject to suspension, the removal of the suspension initially applied, and the expiry of the 30 days without confirmation of the risk profiles, do not preclude the ordinary control powers of the tax administration and, therefore, do not constitute any acknowledgment of the right to the specific tax relief. See Circular No 16/E of 29 November 2021. If the non-existence, even partial, of the requisites that give the right to the tax deduction is ascertained, the tax administration provides for the recovery of the amount corresponding to the deduction not due to the direct beneficiary. The transferee/supplier is jointly and severally liable for the payment of this sum only in the event of concurrence in the violation with wilful deceit or gross negligence. See art. 121, paragraphs 5 and 6, of Law Decree no. 34/2020.
See Revenue Agency Circular No 16/E of 29 November 2021. Therefore, the duty of the transferee/supplier to act with ordinary diligence when they carry out the transaction of transfer or discount on the invoice remains unaffected (does not cease). However, it is believed that, in the event of fraud committed by the transferor/customer, it is first and foremost the responsibility of the tax administration to demonstrate, even with the use of presumptions, the involvement of the transferee/supplier (or according to the well-known «formula» adopted by the Court of Justice for VAT fraud that the transferee/supplier knew or should have known that the transferor/customer was committing fraud) and that only in the face of this proof is the transferee's / supplier’s responsibility to demonstrate that they have acted with ordinary diligence.
Given the limited time frame within which the audit must be carried out, it is unlikely that substantial control will be carried out.
The motivation is provided for by the provision and only for the deed with which the effects of the communication are annulled.
See Rinaldi, R. (2000), Profili ricostruttivi della liquidazione dell’imposta, Lint Editoriale. Coppola, P. (1997), La liquidazione dell’imposta dovuta ed il controllo formale delle dichiarazioni (artt. 36 bis e 36 ter del D.P.R. n. 600/1973). Rassegna tributaria, 14. More recently Zuccarello, A. (2 giugno 2022), Algoritmi e automatismi nei controlli della dichiarazione: profili problematici. Rivista di diritto tributario on line. An automated check is also required for VAT returns. See art. 54 bis of Presidential Decree no. 633 of 1972.
On the implications related to the profiling on the right of privacy v. Scarcella, M. (2019), op. cit.; Drywa, A. (2022), Taxpayer’s Right to Privacy? En Intertax (50, 1), 40-55; Bieker, F. (2022), The Right to Data Protection. Individual and Structural Dimensions of Data Protection in EU Law, Springer. Maja Brkan, M. (2019), The Essence of the Fundamental Rights to Privacy and Data Protection: Finding the Way Through the Maze of the CJEU’s Constitutional Reasoning. German Law Journal (20), 864-883. Kokott, J. — Sobotta, C. (2013), The distinction between privacy and data protection in the jurisprudence of the CJEU and the ECtHR. International Data Privacy Law (Vol. 3, No. 4), 222-228.
In other words, the taxpayer has the right to present written observations and in this way assert their reasons/justifications with respect to the outcome of the automated control.
It should be remembered that with this deed the taxpayer informs the administration of the method of use of the deduction they have chosen.
Fiorentino, S. (2019), Il diritto a una buona amministrazione e la motivazione degli atti tributari. En M. Pierro (coord.), Il diritto ad una buona amministrazione nei procedimenti tributari, p. 245, Giuffrè.
See Moscatelli, M. T. (2005), Il contraddittorio nella fase di liquidazione e controllo formale del tributo alla luce dell’art. 6, comma 5 dello Statuto dei diritti del contribuente. En A. Fantozzi — A. Fedele (coord.), Statuto dei diritti del contribuente, p. 271, Giuffrè. On the right of controverting see Sabbi, L. (2019), Il contraddittorio nel processo tributario, Cedam.
VV.AA. (2019), Il diritto ad una buona amministrazione nei procedimenti tributari, M. Pierro (coord.), Giuffrè.
Council of State, judgment no. 2270 of 8 April 2019, Giornale Dir. Amm., 2019, 6, 781 with a comment by Canalini, V. (2020) L’algoritmo come «atto amministrativo informatico» e il sindacato del giudice; Council of State, judgment no. 8472 of 13 December 2019, Nuova Giur. Civ., 2020, 4, 809 with a comment by Mattera, B. (2020), Decisioni algoritmiche. Il Consiglio di Stato fissa i limiti. Council of State, judgment no. 881 of 4 February 2020, Giur. It., 2020, 7, 1738 with a comment by Orofino, A. G. — Gallone, G. (2020) L’intelligenza artificiale al servizio delle funzioni amministrative: profili problematici e spunti di riflessione. The judgment of the French Conseil Constitutionnel (no. 2018-765 DC of 12 June 2018) can also be placed within this trend, because in which an attempt is made to reconcile the use of algorithmic systems with the possible measures of the public administration.
The Council of State considers in particular that «Né vi sono ragioni di principio, ovvero concrete, per limitare l'utilizzo all'attività amministrativa vincolata piuttosto che discrezionale, entrambe espressione di attività autoritativa svolta nel perseguimento del pubblico interesse» (judgment no. 8472/2019).
According to this case law automation guarantees efficiency and neutrality («efficienza e neutralità») in carrying out administrative activities, with benefits not only in terms of less waste of means and resources and acceleration of the automated process, but also in terms of greater guarantee of impartiality of the automated decision.
Judgment no. 881/2020 exactly states: «elementi di minima garanzia per ogni ipotesi di utilizzo di algoritmi in sede decisoria pubblica»
Judgment no. 881/2020 exactly states: «piena conoscibilità del modulo utilizzato e dei criteri».
Judgment no. 881/2020 exactly states: «imputabilità della decisione all’órgano titolare del potere, il quale deve poter svolgere la necessaria verifica di logicità e legittimità della scelta e degli esiti affidati all’algoritmo «elementi di minima garanzia per ogni ipotesi di utilizzo di algoritmi in sede decisoria pubblica».
Judgment no. 2270 of 8 April 2019 exactly states: «dai suoi autori al procedimento usato per la sua elaborazione, al meccanismo di decisione, comprensivo delle priorità assegnate nella procedura valutativa e decisionale e dei dati selezionati come rilevanti».
Judgment no. 881/2020 exactly states: «necessaria verifica di logicità e legittimità della scelta e degli esiti affidati all’algoritmo».
On the implications of the use of artificial intelligence by public administrations see Galetta, D.U. Algoritmi, procedimento amministrativo e garanzie: brevi riflessioni, anche alla luce degli ultimi arresti giurisprudenziali in materia. En Riv. it. dir. pubbl. com. (3), 501 et seq.; Sola, A. (2020), Inquadramento giuridico degli algoritmi nell’attività amministrativa. En federalismi.it (16), 322 et seq.; Carullo, G.. (2021), Decisione amministrativa e intelligenza artificiale. En Dir. inform. (3), 431 et seq.; Marchetti, B. (2021), La garanzia dello human in the loop alla prova della decisione amministrativa algoritmica. BioLaw Journal (2), 367-385, which analyses the effectiveness of the European Data Protection Regulation.
The Council of State, judgment no. 881/2020, cit., states that «anche l’esercizio di attività discrezionale, in specie tecnica, può in astratto beneficiare delle efficienze e, più in generale, dei vantaggi offerti dagli strumenti stessi».
Due to these shortcomings, human intervention (which is assumed to exist) in the second phase of the procedure could essentially be merely formal. If this is true, then the risk is that the illegitimacy of the administrative measure of suspension will overwhelm the subsequent measure of voiding of the communication, given that the adoption of the latter assumes that the elements of risk are «confirmed». However, this risk is all the more likely given the lack of provision for an intra-procedural right to be heard.
See art. 14 of COM(2021) 206 final, where are indicated the measures which, in the presence of «high-risk» AI systems, must be taken to concretely put the (human) supervisor in a position to decide whether or not to use the algorithmic system.
See Casonato, C. — Marchetti, B. (2021), Prime osservazioni sulla Proposta di Regolamento dell’Unione Europea in materia di intelligenza artificiale. En BioLaw Journal (3), 1 et seq. who talk about «diritto all’umanità nella decisione».
Indeed, the impartiality of the automated decisions depends above all on the impartiality of the construction of the algorithm.
Basilavecchia, M. (2022), L’intelligenza artificiale nell'accertamento, nuovi profili di un vecchio problema. Modulo24 Accertamento e Riscossione. Sole24Ore (7), 66-67, according to which «Gli strumenti tecnologici … omissis … potranno fare grandi cose, ed è auspicabile che le facciano. Ma dev’essere chiaro che anch’essi, non possono togliere centralità al lavoro valutativo dell’Amministrazione né incidere sulla capacità di scelta e di decisione che ad un’Amministrazione fondamentale non può essere né limitata né condizionata, tanto meno da strumenti tecnici che si ritengano a priori più capaci di quanto possa essere l’apporto professionale». De Mita, E. (1991), Interesse fiscale e tutela del contribuente. Le garanzie costituzionali, Giuffrè, 169.
See De Mita, E. (1991), Interesse fiscale e tutela del contribuente. Le garanzie e costituzionali, cit., XV, according to which «l'interesse fiscale non può essere … la comodità del fisco, l'arbitrio, l’irragionevolezza. L’interesse fiscale agisce da parametro per legittimare o meno sul piano costituzionale le ‘particolarità del diritto tributario’ che diversamente non troverebbero giustificazione nell'ordinamento giuridico. La particolarità del diritto tributario … sta … nella predisposizione di regole e istituti funzionali al suo scopo che non consiste nell’imporre comunque somme da pagare, ma nel fare in modo che tali somme pervengano, secondo criteri sostanziali e formali stabiliti dalla legge, nelle casse dello Stato perché possa far fronte alle sue esigenze di bilancio». Fedele, A. (2002), Concorso alle spese pubbliche e diritti individuali. En Rivista di diritto tributario, I, 33 et seq. See also Kuźniacki, B. (11 February 2022), How constitutional principles pave the way to eXplainable AI in tax law. Kluwer International Tax Blog.
In this context we are indeed witnessing a devaluation of the certainty of the law even considered by some to be a legacy of the past. See Pino, G. (2018), La certezza del diritto nello Stato costituzionale. En Diritto pubblico (2), 517. Contra Fabbio, P. (2018), Il dibattito italiano su principi e clausole generali e l’ideale della certezza del diritto. A proposito di un recente libro a cura di Giovanni D’Amico. Osservatorio del diritto civile e commerciale (2), 587-606, according to which the legal certainty «rimane un’istanza per varie ragioni insopprimibile». See also Irti, N. (2016), Per un dialogo sulla «calcolabilità giuridica». En Riv. Dir. Processuale (4-5), 917-925.
The issue of legal certainty has always been the subject of study by tax doctrine primarily due to the poor legislative technique and the continuous issuing of laws. See De Mita, E. (22 January 2017) Quella chimera della certezza del diritto. En Il Sole 24ore. Giovannini, A. (2014), «Certezza del diritto» in materia tributaria: il ruolo della giurisprudenza. Innovazione e diritto (5), 6-16; Zizzo, G. (2022), La clausola generale antiabuso tra certezza del diritto ed equità del prelievo, Pacini Giuridica. On the legislative level, the requests for certainty in the tax field then found express recognition in law no. 212/2000 (known as the «Statuto dei diritti del contribuente»). There are several provisions in the Statuto that have the ultimate aim of strengthening legal certainty in tax matters. For example, there is the provision of non-retroactivity of the tax law required by the second paragraph of art. 3, par. 2 or the one according to which the statute of limitations and forfeiture terms for tax assessments should not be able to be extended (see art. 3, par. 3).
Van Meerbeeck, J. (2016), The Principle of Legal Certainty in the Case Law of the European Court of Justice: From Certainty to Trust. European Law Review (2), 275-288.
Court of Justice of the European Union, 10 March 2009, Gottfried Heinrich, C-345/06, EU:C:2009:140, at 44.
Court of Justice of the European Union, 21 February 2006, C-255/02, Halifax plc v Commissioners of Customs & Excise, EU:C:2006:121, at 76; CJEU, 29 April 2004, C-17/01, Finanzamt Sulingen v Walter Sudholz, EU:C:2004:242, at 34; CJEU, 15 December 1987, C-326/85, Kingdom of the Netherlands v Commission of the European Communities, EU:C:1987:547, at 24.
Court of Justice of the European Union, 5 May 2011, Ze Fu Fleischhandel GmbH (C-201/10) and Vion Trading GmbH (C-202/10) v Hauptzollamt Hamburg-Jonas, EU:C:2011:282.
Court of Justice of the European Union, 16 February 2012, Criminal proceedings against Marcello Costa and Ugo Cifone, C-72/10, EU:C:2012:80, at 74.
Gianformaggio, L. (1988), Certezza del diritto. En Digesto — Discipline privatistiche/sezione civile, Utet. See also Immordino, M. (2003), Il principio di certezza del diritto nei rapporti tra Amministrazione e cittadini, Giappichelli, 24, according to which «esiste una stretta interdipendenza tra certezza del diritto e qualità della normazione, primaria e secondaria».
See articles 22 and 23 GDPR. Regarding this see Scarcella. L. (2019), cit., which stresses «the vagueness» of the safeguards as indicated in the GDPR». In the author's opinion «the presence of these requirements in the law and in the ICT systems effectively used by tax administrations needs to be assessed on a case-by-case basis at the national level. Indeed, the GDPR, by requiring the inclusion of these safeguards, only offers a minimal level of protection that might be extended at the national level».
Slovak Constitutional Court, 17.12.2021, Ústavného súdu Slovenskej republiky PL. ÚS 25/2019-117 V mene Slovenskej republiky, paragraph 122 (text translated with google translate), https://www.slov-lex.sk/pravne-predpisy/SK/ZZ/2021/492/20211217.
De Mita, E. (2017), Quella chimera della certezza del diritto, Il Sole24ore, 22 gennaio 2017.
Ragucci, G. (4 settembre 2019), L’analisi del rischio di evasione in base ai dati dell’archivio dei rapporti con gli intermediari finanziari: prove generali dell’accertamento «algoritmico»? Rivista dir. trib. online.
See art. 1, para. 682-683, law 27 December 2019, no. 160 (2020 Finance Act), which allows the Revenue Agency to use, in performing the risk analysis activities carried out on the basis of the data contained in the archive of financial reports, the technologies, processing and interconnections with the other databases available. The stated aim is to «identify the risk criteria useful for bringing out the positions to be subjected to control and encouraging spontaneous compliance». With the Decree of the Ministry of Economy and Finance of 28 June 2022, the implementing provisions were adopted and the exercise of the rights referred to in the following articles was abridged: 15 («Right of access of the interested party»), 17 («Right to cancellation ("right to be forgotten")»), 18 («Right to limitation of treatment») and 21 («Right to object») of regulation (EU) no. 2016/679.