Use of Artificial Intelligence in Detecting Tax Non-Compliance Risks: A Policy Assessment with a Focus on Hybrid Mismatch Arrangements

This article explores artificial intelligence (AI) technologies as an effective tool for tax administrations to identify non-compliance risks, using hybrid mismatch arrangements – a form of tax avoidance difficult to detect through traditional audit techniques – as a case study. It aims to demonstrate how algorithmic capabilities can strengthen compliance risk management and audit case selection within legal and governance constraints. The study examines specific techniques, including machine-readable/consumable legal texts, social network analysis, graph analytics, genetic algorithms, and natural language processing (NLP), to flag patterns consistent with deduction/no inclusion (D/NI), double deduction (DD), dual residence, imported mismatches, and structured arrangements. Furthermore, it analyses the legal and institutional dimensions of AI adoption, drawing on country practices and the literature. The article argues that aligning AI with legal frameworks, specifically Base Erosion and Profit Shifting (BEPS) Action 2 and Anti-Tax Avoidance Directive (ATAD) 2 linking rules, under transparent governance, enhances audit targeting and prioritization. Ultimately, the study highlights that the accountable use of AI ensures a consistent application of linking-rule analyses, fostering the transition toward risk-based, data-driven, and transparent tax administrations.