Governance of algorithmic regulation in Brazil: pathways, principles, and challenges for Public Administration

Abstract

This paper analyzes the legal and institutional challenges of regulating artificial intelligence in the Brazilian public sector, with a focus on administrative action and regulatory governance. It begins by addressing the growing presence of AI in public services and the associated risks of opacity, algorithmic discretion, and regulatory capture. The study then explores the theoretical foundations of algorithmic regulation, distinguishing between regulation by algorithms and regulation of algorithms. It critically examines the inadequacy of the current Brazilian legal framework, having into account the Bill 2.338/2023, in addressing the challenges posed by algorithmic administration, such as the duty to provide reasoning and the explainability of automated decisions. A comparative legal analysis is conducted with models from the European Union, Canada and Germany, identifying regulatory best practices while also highlighting the challenges of directly transposing such frameworks into the Brazilian context. The research adopts a hypothetical-deductive method, based on the hypothesis that Bill 2.338/2023 fails on providing effective regulation of the issue. The conclusion takes in consideration that building an effective regulatory framework for AI in public administration requires not only normative instruments, but also robust institutions, an accountability culture and a commitment to constitutional principles of legality, transparency and control.