PREDICTIVE POLICING, CRIMINAL JURISDICTION AND ARTIFICIAL INTELLIGENCE: THE ROLE OF THE PUBLIC PROSECUTION SERVICE

Abstract

Contextualization: The article analyzes the extent to which the use of predictive policing tools and artificial intelligence techniques by the Public Prosecution Service, as a basis for forming the opinio delicti and for producing criminal evidence, is compatible with constitutional and procedural safeguards within the criminal jurisdiction in Brazil. It starts from the expansion of algorithmic technologies and statistical models oriented towards risk management, as well as from the growing diffusion of actuarial justice rationality in criminal proceedings.

Objectives: To examine how predictive policing, especially in its person-based variant, strains classic categories of a rights-oriented criminal procedure – such as the presumption of innocence, adversarial proceedings, reasoned decision-making, protection of personal data and judicial control over evidence – and to delimit the constitutional role of the Public Prosecution Service, within criminal jurisdiction, between reproducing and containing this actuarial rationality.

Method: The study adopts a legal-dogmatic and critical-prospective approach, combining the analysis of the 1988 Constitution, infra-constitutional legislation, and international human rights standards with critical criminology and with national and foreign literature on algorithms, predictive policing, actuarial justice, artificial intelligence, and algorithmic discrimination.

Results: The article shows that the opacity of predictive systems, the probabilistic nature of risk scores, and the absence of a robust algorithmic chain of custody can intensify selective and discriminatory practices, hinder judicial control over evidence, and weaken the exercise of the right to defence in criminal jurisdiction. At the same time, it acknowledges the possibility of targeted uses compatible with fundamental rights, such as supporting the identification and location of fugitives by cross-referencing databases and using automated systems, provided that such tools are treated only as investigative inputs and subjected to strict filters of legality, proportionality, and non-discrimination by the Public Prosecution Service.