The 2024-2027 Strategic Plan establishes the principles that should guide the use of Artificial Intelligence for the prevention and fight against tax fraud. However, there is a lack of express references to taxpayers’ rights and principles (such as transparency and proportionality, or the principle of non-discrimination) that are widely recognized and whose protection is essential in this area.
The new Strategic Plan of the Tax Agency 2024-2027 announces a significant effort in the use of Artificial Intelligence (“AI“) for the prevention and fight against tax fraud. Although the use of AI by the AEAT is not necessarily new (references to its use can already be found, among other documents, in the “Strategic Plan of the Tax Agency 2020-2023” and in the “National Strategy of Artificial Intelligence of Spain” presented by the Government in 2020), in this new strategic plan, the Public Treasury’s commitment to AI can be seen with greater intensity, which allows us to anticipate important changes in its future relationship with taxpayers.
AI is defined in the strategic plan as the set of systems capable of functioning with certain levels of autonomy to achieve objectives and generate information in the form of predictions, recommendations, or decisions based on data, applying learning strategies that are automatic or based on logic and knowledge.
According to this definition, the AI referred to by the AEAT consists of a system that will be able to automatically (or almost automatically) use the available machine learning / deep learning tools to draw from the abundant sources of information available to the Tax Administration (big data) and make predictions, recommendations, or even make decisions that affect taxpayers. In other words, this AI goes far beyond the classic modelling, automation, and “data crossing” tools that the AEAT has traditionally had and is presented as a system that will be able to “reason” and analyse data autonomously to make its own decisions in its relationship with the taxpayer (regardless of the fact that the final decision-making continues to fall, as it cannot be otherwise, on people).
On the other hand, the objective of the use of AI by the AEAT, as underlined in the plan, lies in the need to use (in a safe way) innovation and emerging technologies to improve the effectiveness and efficiency of tax and customs actions (i) in matters of information and assistance to the taxpayer and (ii) in the field of prevention and fight against tax and customs fraud.
In the development of these objectives, the AEAT expressly assumes the following commitments:
- Adjust the use of AI so that it is in line with the vision, mission, values, and goals of the Tax Agency, ensuring that its development adheres to current legislation.
- Employ AI with the aim of improving the services offered to citizens and increasing the efficiency and effectiveness of administrative management, guaranteeing respect for the principles of equity, objectivity, and uniformity in the operations of the Tax Agency.
- Explore and advance the application of AI to support compliance with principles such as rationalization, economy, and efficiency, which are fundamental for any public body.
In addition, the following guiding principles are established in the use of AI by the AEAT: (i) respect for the “principle of responsibility”, which (although not defined in the document) requires that responsibilities and functions in the use of AI are correctly assigned and defined and obliges the AEAT to be accountable for any negative impact derived from its application; (ii) the adoption of a Human Centric approach, to guarantee the safe and ethical use of AI (providing protection of citizens’ rights and the ethical principles of the AEAT); and, finally, (iii) attention to the “principle of security and AI governance” both in the AI systems themselves and in taxpayer data (which must be safeguarded with the highest security guarantees).
The plan states that AI projects will be carried out by teams composed of experts in various areas, including specialized officials, computer technicians, and data scientists, supervised by the AEAT’s Department of Tax Informatics. In this specific aspect, the AEAT ensures the use of quality data and the permanent and complete supervision of the systems by human beings.
Finally, it is indicated, in what could be interpreted as a warning to taxpayers, that the limitation to the use of AI by the AEAT will only be determined by the legal-normative framework applicable at any given time and by the guiding principles of AI and the values of the AEAT. However, it is established that human intervention in final decision-making will be guaranteed.
In short, the Strategic Plan 24-27 announces a firm and determined will of the AEAT to use the AI tools available at any given time to control the relationship it maintains with taxpayers, although with the commitment to respect (i) the “principles of good use of AI” of the AEAT itself (which, in any case, are scarcely defined, and further development would be desirable, as a guarantee of taxpayers’ rights), and (ii) the legal-normative framework that may be applicable at any given time (which, in terms of AI, does not yet have sufficient development).
We must applaud, without a doubt, that the AEAT itself recognizes the existence of certain guiding principles that, as a form of internal self-control, should govern its use of AI tools in its relationship with the taxpayer. However, there is a lack of express mention of other guiding principles and rights of citizens that, in our opinion, are already widely accepted and that should also have been expressly recognized in this document. By way of example, the following can be cited:
- a) The “principle of transparency” together with the consequent “right of explanation”, by which the taxpayer should have access to the technical aspects, algorithms, and data used by these AI systems for their decision-making (with the logical restrictions essential to ensure their effectiveness as a mechanism for the fight against tax fraud).
- b) The “principle of proportionality”, which guarantees that AI tools are used only for the specific purposes for which they were created and that the necessary balance that must exist between the public purpose that justifies their use (consisting of the prevention and fight against tax fraud) and the rights of citizens to freedom, privacy, and legal certainty is assessed at all times.
- c) The “principle of equity, non-discrimination, and absence of bias” in the relationship of the AEAT with taxpayers, which guarantees that automated decisions based on issues such as sex, race, religion, social class, or ideologies are not made by AI, even if this may restrict the effectiveness of these systems.
- d) The taxpayer’s “right to challenge” decisions based on AI that they consider incorrect or unfair.
Expanding on this, under the hypothesis of the need to use technology to comply with the objectives of preventing and combating tax fraud and praising the self-regulation that the AEAT is voluntarily developing in this matter through the Strategic Plan 2024-2027, we must advocate for the development in Spain of a legal framework, of mandatory compliance and with the rank of Law, that specifically regulates the use of AI by the AEAT, in which the guiding principles and rights of citizens in this area are defined in a complete manner and in protection of taxpayers’ rights.
