AUDITABILIDADE ALGORÍTMICA E DEVIDO PROCESSO

O HABEAS CODE COMO CATEGORIA DOGMÁTICO-PROCEDIMENTAL DE CONTESTAÇÃO DE DECISÕES AUTOMATIZADAS

Abstract

This article examines the insufficiency of the legal and procedural instruments currently available under Brazilian law to ensure the contestability of automated decisions that affect fundamental rights. It starts from the premise that contemporary algorithmic harm does not always arise from inaccurate or unlawfully stored personal data, but may result from the processing logic that transforms data into classifications, profiles, scores, account blocks, exclusions, or restrictions. The research adopts a qualitative legal-dogmatic and legal-propositional approach, based on bibliographic review, normative analysis, and critical reconstruction of the limits of habeas data, the Brazilian General Data Protection Law, mandamus-like remedies, and ordinary evidentiary techniques. It argues that automated decisions with significant impact are compatible with due process only when they provide minimum conditions of functional auditability, without requiring absolute transparency, unrestricted access to source code, or the suppression of trade secrecy. The article proposes to understand Habeas Code not as a new autonomous constitutional action, but as a dogmatic-procedural category aimed at organizing, through existing legal instruments, mandatory relief for algorithmic auditability. It concludes that this category may contribute to structuring duties of individualized explanation, preservation of technical records, controlled access to decision-making logic, technical adversarial procedure, substantive human review, and proportionate remedies for the correction, suspension, or invalidation of opaque automated decisions.