Review Capacity and Contestability: Procedural Justice Beyond the EU AI Act in Technologically Dependent States

Abstract

This article examines whether the procedural safeguards of the European Union Artificial Intelligence Act can operate effectively in technologically dependent states. Using doctrinal analysis of the Act and comparative public-law materials, it argues that transparency, documentation, logging, human oversight, impact assessment and monitoring presuppose a prior institutional condition: review capacity. Where public authorities depend on external vendors to design, operate, maintain or audit AI systems, legal responsibility may remain with the state while practical control lies elsewhere. The resulting accountability gap is examined through the concept of review dependency and through litigation concerning automated welfare administration and a hypothetical digital-identity example. The article argues that inadequate review capacity may produce a de facto ex parte form of administration, in which affected persons cannot meaningfully answer the technical assessment underlying a decision. Drawing on audi alteram partem, it proposes a risk-calibrated framework of notice, explanation, competent human review, contestability and institutional supervision. Informed, withdrawable consent has a role where participation is genuinely voluntary; mandatory public functions require effective procedural safeguards independent of individual consent. The central conclusion is that adopting EU-style safeguards requires corresponding legal, technical and institutional capacity to make review effective.