Open Access
ARTICLE
Artificial Intelligence in Tax Administration: Legal Limits and Regulatory Risks
Issue Vol. 2 No. 01 (2026): Jan Feb 2026 --- Section Articles --- Published Date: 2026-02-28
Abstract
The accelerating deployment of artificial intelligence and machine learning systems in tax administration is reshaping the relationship between fiscal authorities and taxpayers in a manner that existing legal frameworks are ill-equipped to address. Automated risk scoring platforms, predictive audit selection tools, and large-scale data-matching infrastructures now exercise profound influence over enforcement outcomes in jurisdictions across the European Union, the United States, the United Kingdom, and Australia, yet they operate in conditions of structural opacity that are incompatible with foundational requirements of due process, equal treatment, and effective judicial review. This article examines the legal limits and regulatory risks inherent in this transformation, drawing upon comparative constitutional and administrative law, European Union data protection and artificial intelligence regulation, and the emerging scholarship of algorithmic governance. Central to the analysis is the Dutch Toeslagenaffaire - the childcare benefits scandal in which a machine-learning algorithm operated by the Belastingdienst produced discriminatory outcomes affecting approximately 35,000 families, ultimately causing the resignation of the Rutte II cabinet - as a paradigm case that exposes the systemic legal deficits of algorithmic enforcement. The article critically assesses the adequacy of the EU AI Act (Regulation (EU) 2024/1689), the General Data Protection Regulation, and OECD soft-law instruments, argues that all three are structurally deficient as applied to tax administration AI, and advances a framework of de lege ferenda reform proposals grounded in legality, explainability, human oversight, non-discrimination monitoring, and effective remedy.
Article Cover Image
