In recent years, the Tax Administration of the Province of Córdoba – a subnational jurisdiction of the Argentine Republic – has progressively incorporated a specific regulatory framework for the use of artificial intelligence (AI) technologies in the tax field, with the aim of integrating them as a concrete, regulated and legally structured reality.

In a previous publication of our co-authorship, disseminated by the CIAT[1], we analyzed how the Province of Córdoba made a relevant first step by recognizing – in an express and normative way – the use of AI within the tax procedure, accompanying it with explicit guarantees for taxpayers and /or responsible, as well as self-imposed obligations to the Administration itself. That experience demonstrated that technological innovation does not necessarily conflict with fundamental rights, provided that there is a clear and precise legal framework to guide it.

The amendments to the Provincial Tax Code  in force from 2026 – introduced by Law No. 11,089 – now allow us to take one more step. It is no longer just a matter of enabling the use of AI under certain conditions, but of institutionalizing its presence, defining responsibilities, establishing external controls, and, at the same time, precisely delimiting what the technology can do and what remains, in a non-delegable way, a human function.

The purpose of this document is to inform other Tax Administrations of the regulatory advances that have been made, which may be useful for those jurisdictions that, within their objectives, plan to coexist with intelligent systems in the field of tax law.

External audits: AI is no longer a “black box”

One of the most relevant changes introduced by the reform is to require that artificial intelligence systems used in verification and/or control processes can be subjected to external and independent audits[2].

If AI is going to influence – even indirectly – decisions with fiscal impact, its operation cannot be circumscribed and/or limited to a purely internal area of administration. The possibility of universities, research institutes, or specialized public bodies evaluating these systems introduces a standard of control that goes beyond the merely formal.

From the perspective of the taxpayer and/or responsible, this forecast reinforces basic guarantees. It is not a question of accessing the source code of the systems used, but of ensuring that the models applied and/or used are reliable, impartial and reasonably free of discriminatory biases, avoiding that the technology operates as an element lacking transparency and/or verifiable control within the tax procedure.

For the tax administration itself, far from being a burden, this scheme would offer relevant institutional support. An audited, evaluated and/or documented system would reduce potential litigation and strengthen the reasonableness of tax action in the face of possible administrative and/or judicial challenges.

The standard does not detail specific methodologies or technical scopes, which leaves a necessary margin of flexibility in a constantly evolving field. But it does set a guiding principle, in the sense that the AI operating in the tax field must be able to be explained, controlled and evaluated by suitable third parties.

Technology delegation: the responsibility remains with the “Taxpayer”

Another central aspect of the reform is the express incorporation of the concept of responsibility by technological delegation. The Tax Code recognizes an increasingly widespread reality: taxpayers and/or responsible persons use automated systems, algorithms and/or intelligent platforms to settle, declare and fulfill their tax obligations[3].

Faced with this scenario, the regulation adopts a classic but necessary position in relation to the fact that tax liability is not delegated to technology. Decisions made by automated systems are directly attributed to the taxpayer and/or responsible party, even when they have been executed by an algorithm.

This approach excludes the possibility of transferring and/or mitigating responsibilities on the grounds that “the system failed”. The decision to automate processes implies assuming a technological risk that does not shift the ownership of the tax obligation.

However, the reform introduces a relevant nuance. When the breach derives directly from the design, configuration or operation of the system, and the intent or fault of the supplier or developer is credited, the normative enables the joint and several liability of those third parties. It is not a question of an objective or automatic responsibility, but of a carefully delimited scheme, supported by traditional criteria of subjective imputation.

In addition, the administration is empowered to request technical documentation and / or operational records of the system, always under the protection of confidentiality and intellectual property. This point is key to reconstructing the traceability of automated decisions and delineating responsibilities in complex technological environments.

In short, AI can assist, but it does not replace either the taxpayer or his legal responsibility.

Predictive tax communication: prevention without penalties

The reform is also advancing on a less conflictive, but no less relevant field: AI-based predictive tax communication. The Code enables the administration to use intelligent systems to anticipate maturities, detect inconsistencies or warn of default risks, through various digital channels[4].

These communications are of a purely informative nature, do not produce legal effects or interrupt administrative deadlines. They are not administrative acts and do not generate new obligations. They are, in simple terms, warnings.

This regulatory design avoids frequent confusion in digital environments and preserves legal certainty. AI is used as a prevention and support tool, not as a covert intimation and/or control mechanism.

From a tax policy perspective, these tools make it possible to reduce involuntary defaults, improve the tax-taxpayer relationship and reduce administrative costs, without altering the legal position of the parties.

This consolidates an idea introduced by the reform: technology can help, alert and assist, but not decide or sanction on its own.

Online information regime: data, control and limits

An aspect that deserves a specific reference within the new regulatory change is the incorporation of the mandatory and online information regime, provided for in Article 11 bis of the Provincial Tax Code[5]. Although this is not a rule expressly referring to AI, its analysis allows us to warn that it would play a structural role within the technological ecosystem that the Province has been building, particularly in the context of an increasingly digitalized economy.

The faculty granted to the Ministry of Public Revenue to establish information regimes in real time – or in technically equivalent operational periods – responds to a specific need, in the sense of providing the Tax Administration with timely, standardized and verifiable data, especially in contexts where economic operations take place in digital, delocalized environments and of difficult territorial apprehension.

From this perspective, the aforementioned article functions as an important piece of legislation that would allow feeding the analysis, verification and /or control systems that, eventually, may incorporate risk profiling mechanisms assisted by AI tools. Without data, there is no real possibility of AI application.

The relevance of the precept is not in the breadth of the faculty granted, but in the express limits that the normative imposes. The information regime is strictly limited to operational data essential to verify the existence of taxable transactions in the field of Gross Income Tax and the due compliance with tax obligations, excluding any access to content, private communications, consumption profiles, commercial preferences or individualized traceability of users.

This point is not minor. In this sense, the provincial legislature authorizes the necessary information and prohibits invasive information. The regulation does not limit itself to stating generic data protection principles, but expressly lists what is beyond the scope of the treasury, even if it is technically accessible.

The incorporation of objective presumptions regarding the location of the taxable event, based on elements such as IP geolocation, the use of payment methods, and/or telephone lines associated with the province, is also significant. These presumptions, which are accepted unless proven otherwise, make it possible to resolve—using reasonable and verifiable criteria—one of the main challenges of taxation in digital environments, without resorting to broad fictions or invasive reconstructions of user behavior.

From a data governance perspective, the article reinforces criteria of necessity, adequacy, and proportionality, requires security and encryption protocols, and even enables the use of pseudonymization techniques when personal data is not essential. This regulatory design directly addresses the principles governing the responsible use of artificial intelligence, data minimization, traceability, specific purpose, and control of processing.

In short, the planned information regime cannot be read as a simple extended control tool. Rather, it is a legal framework that seeks to balance access to information with the protection of rights, and which is essential for any future use of AI in tax matters to be based on data obtained in a legitimate, secure and legally controllable way.

Article 11 bis consolidates the idea of a technically informed, but legally limited treasury, a condition that would be necessary so that technological innovation does not lead to undue asymmetries or affect the privacy of taxpayers and/or users.

Conclusion.

The amendments introduced to the Tax Code of Córdoba consolidate a coherent regulatory framework for the incorporation of artificial intelligence in the tax field. In the first stage, duly analyzed in the CIAT, the Province normatively recognized the use of AI as a tool for taxpayer assistance and support in verification and / or control processes, establishing human intervention as a non-delegable principle in every decision-making instance.

The reforms subsequently introduced by Law No. 11,089 – effective from 2026 – consolidate this process, by deepening this framework through the provision of external audits, mandatory human control, the precise allocation of responsibilities and a clear differentiation between informative and decision-making functions, allowing technology to be integrated without affecting the classic principles of tax law.

The Cordovan experience shows that technological innovation is sustainable when it is based on clear rules and defined limits, reaffirming that artificial intelligence does not replace tax law, but assists it, always under legality, human control and institutional responsibility.

References:

[1] https://www.ciat.org/ciatblog-inteligencia-artificial-y-derechos-de-los-contribuyentes-el-caso-innovador-de-la-provincia-de-cordoba/.

[2] Article 17 -relevant part- “v) To guarantee that the artificial intelligence and/or risk profiling systems used in the verification and/or control processes can be subject to external and independent audits through public or private universities, research institutes belonging to the public scientific and technological system and/or by public bodies with technical competence in the field, in order to evaluate their reliability, impartiality and absence of discriminatory biases. The resulting reports may be published on an annual basis, respecting the rules of tax secrecy and the protection of personal data.”

[3]  “Tax Liability for Technological Delegation.

Article 40 bis.- The subjects liable for the taxes provided for in this Code and in the other tax regulations of the Province that use automated systems, algorithms or artificial intelligence as means or platforms for the settlement, declaration, payment or fulfillment of tax obligations, will be responsible for the tax that derives from them – even if it is a technological delegation -, the decisions taken by such systems being directly imputed to the taxpayer.

When non-compliance with a tax obligation arises directly and immediately from the operation, design, or configuration of the automated system, and intent or negligence on the part of the supplier, developer, or administrator of the system is proven, they may be held jointly and severally liable to the extent of their active participation in the implementation, maintenance, or technical advice of the system.

Tax authorities, with the due guarantees of confidentiality and protection of intellectual property, may require the technical documentation, the operation records and the system configuration criteria, in order to verify its regulatory adequacy, the traceability of automated tax decisions and the possible liability of the intervening third parties.”

[4] “Multichannel Predictive Tax Communication.

Article 46 bis.- The authority may implement an informative predictive tax communication system, based on artificial intelligence technologies, that allows anticipating deadlines, inconsistencies, omissions or risks of non-compliance by taxpayers and/or responsible parties.

Communications may be made by multichannel digital means, including email, mobile applications, web platforms, official social networks, virtual assistants and any other channel validated by the Authority, provided that traceability, authenticity and effective reception are guaranteed.

The communications issued within the framework of this article will not produce legal effects or interrupt administrative deadlines, having only the value of a notice or courtesy reminder.”

[5] “Mandatory Information System.

Article 11 bis.- Empower the Secretariat of Public Revenue, under the Ministry of Economy and Public Authority – or the body that will replace it in the future – to establish, through the forms, deadlines and conditions that it has, a mandatory online information regime, which allows the Directorate to access in real time, or in technically defined equivalent operational periods, the necessary data for tax verification, regarding taxpayers and / or managers, residents or not in the province of Córdoba, who market goods and / or provide services, digital or non-digital, to users and / or consumers domiciled, based or functionally linked with the province.

The regime must be limited exclusively to operational data essential to verify the existence of operations achieved by the taxes legislated in this Code and other provincial tax regulations, as well as the proper fulfillment of tax obligations, being prohibited the request for information that implies access to contents, private communications, consumer profiles, commercial preferences, individualized traceability of the user or any data unrelated to the fiscal purpose.

For the purposes of this article, it will be presumed, unless proven otherwise, that the user and/ or consumer is located in the province when the operation is carried out through an IP address geolocated in its territory, a SIM card or telephone line associated with the province is used, a means of payment issued or registered in it intervenes, or other objective and verifiable elements allow to reasonably infer the location of the taxable event of the taxes legislated by this Code and by the other provincial tax regulations.

The information that the Secretariat has must be requested and transmitted with criteria of strict necessity, adequacy and proportionality, using security, confidentiality and encryption protocols, and pseudonymization techniques may be applied with respect to those personal data of the user or consumer that are not indispensable to identify, individualize or locate the operation for the purposes of its fiscal verification, without affecting in any case the possibility of verification, supervision and / or determination of the tax obligation by the Authority.

The information will be used exclusively for tax purposes and in no case may it be transferred, shared or processed for purposes other than those provided for in this Code, nor give rise to the processing or reconstruction of sensitive personal data or financial information of a reserved nature.

The Ministry of Public Revenue may conclude agreements with provincial agencies competent in information technologies in order to ensure the proper management, safeguarding, integrity, availability and traceability of the data obtained, as well as apply proportional and gradual measures in the event of non-compliance with the regime established, in accordance with the provisions of this Code.”

25 total views, 25 views today