A.I. in the Italian criminal justice system in the light of the constitutional principles: potentialities, risks and perspectives

The research aims to verify if and how A.I. is compatible with the constitutional principles of the Italian criminal justice system and the european legislation. While on the pre-investigative level, for predictive policing activities there are X-law and KeyCrime softwares, on the trial level the use of A.I. still remains hypothetical. Profiling techniques, risk assessment tools, assessment of eyewitness’s evidence tools, while promising prodigious results in reconstructing the truth, seem to betray the principles of due process. In fact, the inaccessibility of the source code and the self-learning system make the algorithmic response opaque: firstly, for the judge, who will not be able to explain the reliability of the algorithmic recommendation; then, for the parties, unable to challenge and falsify the output, with serious injury to the rights of defense and of cross-examination. Yet, the prospects of A.I. on the performance of judicial offices appear to be constitutionally sustainable: new "intelligent" databases of case law and doctrine, but also A.I. systems to optimize time and resources, assign files, and schedule hearings. Despite the registered perplexities and the need for "meaningful human control," this is an opportunity to be grasped, also in view of the Recovery Plan.