The use of AI in judicial systems: From documentary assistance to jurisdictional management

Digital transformation brings artificial intelligence (AI) tools to the judicial ecosystem with a wide range of uses: from documentary assistance (search, classification, summarization, anonymization) to support for jurisdictional management (planning, resource allocation, administrative sorting) and, at the limit, to analyses that can influence the evaluation of evidence and the outcome of cases. In this context, the article proposes an integrated framework for judicial governance of AI, aligned with the EU Regulation on AI (AI Act), which combines risk classification, AI literacy of staff, audit (logging, traceability) and explainability of automated decisions, with the explicit aim of maintaining the legitimacy of human decision-making. From a methodological perspective, the paper maps the judicial uses of AI on the risk classification of the AI Act and extracts the related obligations (transparency, human oversight, risk assessment and mitigation, event logs), with a focus on the first provisions already operational (AI literacy and prohibited practices). The central contribution is a model of “judicial AI governance” that includes: registry of AI uses and risk matrix; algorithmic impact assessments; policies of journaling and explainability appropriate to the target audience (judges/registrars/parties); mechanisms of challenge and redress; joint legal-technical-ethical committee; role-based AI literacy programs. The article derives sets of indicators (rate of human interventions, quality of motivation, incidents of bias/challenge) and draws guidelines for integrating AI without compromising procedural guarantees (adversarial, publicity, right to defense). The conclusion is that the legitimacy of human decision does not result from the absence of AI, but from its design and governance in such a way that control and responsibility remain, verifiably, with the human judge.