Governing AI in Tax Administration: A Comparative Constitutional Framework for Legality, Transparency, Oversight, and Redress

Tax authorities increasingly deploy algorithmic and AI-assisted systems for audit selection, fraud detection, and automated penalties, yet the legal conditions for their use remain underspecified. This article develops a comparative constitutional framework for algorithmic tax governance, drawing on UK statutory and case law, the European Convention on Human Rights (ECHR), the EU General Data Protection Regulation (GDPR), the EU Artificial Intelligence Act, and recent Dutch jurisprudence. It identifies three governance models – the Dutch legality model, the ECtHR proportionality model, and a UK delegative-transparency model – and argues that all three converge on four minimum conditions: explicit legal authority, intelligible transparency, meaningful human oversight, and effective redress. The framework is a portable diagnostic instrument for jurisdictions – including Australia, Canada, Ireland, and EU Member States – relying on broad statutory delegation with fragmented ex post safeguards. Section 103 of the Finance Act (FA) 2020 authorizes automation but does not supply those conditions, leaving safeguards fragmented across tax procedure, data-protection law, and information-law litigation. The article concludes that algorithmic tax administration requires ex ante legislative governance rather than ex post reform.