Towards Responsible AI Governance in the Brazilian Judiciary

Artificial intelligence (AI) has been increasingly adopted in judicial systems, including in Brazil, where over 100 AI tools have been implemented since the early 2020s under the National Council of Justice’s (CNJ) Justice 4.0 program. While these systems such as Victor and Athos support legal document classification, case triage, and jurisprudence recommendation, their use raises concerns about transparency, fairness, and alignment with constitutional and legal norms. This research examines how responsible AI is being operationalized in the Brazilian judiciary, analyzing the legal frameworks, governance strategies, and institutional infrastructures guiding its deployment. Drawing on regulatory developments including CNJ Resolutions No. 332/2020 and 615/2025 and a national CNJ survey of 140 AI models, the study identifies both advances and tensions in areas such as transparency, human oversight, explainability, and ethical risk management. Through document analysis and semi-structured interviews with judges and judicial administrators, it explores how principles like justice and accountability are translated into practice, contributing to broader debates on AI governance and the role of automation in public institutions. Initial findings indicate a shift from a principle-based framework in 2020 to a more prescriptive and detailed governance model in 2025, incorporating risk classification, stronger transparency requirements, and specific rules for generative AI, while raising questions about potential challenges in the practical enforcement of governance measures.