The widespread dissemination of big data and artificial intelligence systems has profoundly affected social, economic, and institutional dynamics, extending to public decision-making processes and, in particular, to the functioning of criminal justice. Machine learning technologies, capable of analysing vast amounts of data and generating predictive outputs, are already being employed as tools to support judicial activity, especially in specific sectors and legal systems, with the aim of enhancing the efficiency, speed, and consistency of decision-making processes. Their use is likely to intensify and expand, confronting the law with unprecedented challenges. This study examines the relationship between artificial intelligence and criminal procedure, with particular emphasis on the evidentiary domain, where the impact of new technologies is most evident. After outlining the relevant conceptual and regulatory framework, the analysis explores the principal applications of AI within criminal proceedings, distinguishing between tools for organisational and administrative support, investigative technologies, and systems capable of directly influencing the formation and evaluation of evidence. Specific attention is devoted to the phenomenon of machine-generated evidence, both from an ex post perspective—concerning the reconstruction of events that have already occurred—and from an ex ante perspective, relating to the use of predictive algorithms to formulate prognostic judgments, particularly through risk assessment tools employed to estimate the risk of recidivism. The analysis highlights that, while such applications may expand the judge’s epistemic resources and provide support in fact-finding, they also raise significant concerns in terms of transparency, verifiability, accessibility of algorithmic evidence, and respect for the adversarial principle. The opacity of decision-making models, the risk of bias and discrimination, the difficulties associated with reviewing automated decisions, as well as the potential weakening of judicial independence, call for careful reflection on the limits within which the legal system may legitimately rely on such instruments. The research proceeds from the assumption that the interaction between artificial intelligence and criminal procedure is now inevitable and already underway. Rather than questioning the mere appropriateness of its use, the study seeks to identify conditions, safeguards, and regulatory models capable of integrating new technologies into the procedural system in a manner consistent with the epistemological principles of criminal proceedings and the protection of fundamental rights. From this perspective, the paper advocates the emergence of a new phase in criminal procedure, grounded in human–machine collaboration, capable of reconciling technological progress, the reliability of tools, and the safeguarding of the principles of a fair trial, within a clear and precise regulatory framework.
La diffusione dei big data e dei sistemi di intelligenza artificiale ha inciso in modo profondo sulle dinamiche sociali, economiche e istituzionali, fino a coinvolgere anche i processi decisionali pubblici e, in particolare, il funzionamento della giustizia penale. Le tecnologie di apprendimento automatico, capaci di analizzare grandi moli di dati e di generare output predittivi, sono già oggi impiegate come strumenti di supporto all’attività giudiziaria, in particolare in specifici settori e ordinamenti, con l’obiettivo di incrementare l’efficienza, la rapidità e la coerenza dei processi decisionali. Il loro utilizzo appare destinato a intensificarsi ed espandersi, ponendo il diritto di fronte a sfide inedite. Il lavoro analizza il rapporto tra intelligenza artificiale e processo penale, con particolare attenzione all’ambito probatorio, nel quale l’impatto delle nuove tecnologie risulta maggiormente evidente. Dopo aver delineato il quadro concettuale e normativo di riferimento, lo studio esamina le principali applicazioni dell’IA nel procedimento penale, distinguendo tra strumenti di supporto organizzativo e amministrativo, tecnologie investigative e sistemi capaci di incidere direttamente sulla formazione e sulla valutazione della prova. Un focus specifico è dedicato al fenomeno della machine-generated evidence, sia nella prospettiva ex post, relativa alla ricostruzione di fatti già verificatisi, sia in quella ex ante, concernente l’impiego di algoritmi predittivi per la formulazione di giudizi prognostici, in particolare attraverso i risk assessment tools utilizzati per stimare il rischio di recidiva. L’analisi mette in luce come tali applicazioni, pur offrendo un potenziale ampliamento del patrimonio conoscitivo del giudice e un supporto all’accertamento dei fatti, sollevino rilevanti criticità in termini di trasparenza, verificabilità, accessibilità della prova algoritmica e rispetto del contraddittorio. L’opacità dei modelli decisionali, il rischio di bias e discriminazioni, le difficoltà di controllo delle decisioni automatizzate, nonché il possibile indebolimento dell’indipendenza giudiziaria impongono una riflessione attenta sui limiti entro cui l’ordinamento può legittimamente avvalersi di tali strumenti. La ricerca muove dall’assunto che il contatto tra intelligenza artificiale e processo penale sia ormai inevitabile e già in atto. Piuttosto che interrogarsi sulla mera opportunità del suo impiego, il lavoro si propone di individuare condizioni, garanzie e modelli regolatori idonei a integrare le nuove tecnologie nel sistema processuale in modo coerente con i principi epistemologici del processo e con la tutela dei diritti fondamentali. In questa prospettiva, si auspica l’avvio di una nuova stagione del processo penale, fondata sulla collaborazione tra uomo e macchina, capace di coniugare progresso tecnologico, affidabilità degli strumenti e salvaguardia dei principi del giusto processo, all’interno di una cornice normativa chiara e puntuale.
L'intelligenza artificiale per la giustizia penale
AMIDANI, DIEGO
2026
Abstract
The widespread dissemination of big data and artificial intelligence systems has profoundly affected social, economic, and institutional dynamics, extending to public decision-making processes and, in particular, to the functioning of criminal justice. Machine learning technologies, capable of analysing vast amounts of data and generating predictive outputs, are already being employed as tools to support judicial activity, especially in specific sectors and legal systems, with the aim of enhancing the efficiency, speed, and consistency of decision-making processes. Their use is likely to intensify and expand, confronting the law with unprecedented challenges. This study examines the relationship between artificial intelligence and criminal procedure, with particular emphasis on the evidentiary domain, where the impact of new technologies is most evident. After outlining the relevant conceptual and regulatory framework, the analysis explores the principal applications of AI within criminal proceedings, distinguishing between tools for organisational and administrative support, investigative technologies, and systems capable of directly influencing the formation and evaluation of evidence. Specific attention is devoted to the phenomenon of machine-generated evidence, both from an ex post perspective—concerning the reconstruction of events that have already occurred—and from an ex ante perspective, relating to the use of predictive algorithms to formulate prognostic judgments, particularly through risk assessment tools employed to estimate the risk of recidivism. The analysis highlights that, while such applications may expand the judge’s epistemic resources and provide support in fact-finding, they also raise significant concerns in terms of transparency, verifiability, accessibility of algorithmic evidence, and respect for the adversarial principle. The opacity of decision-making models, the risk of bias and discrimination, the difficulties associated with reviewing automated decisions, as well as the potential weakening of judicial independence, call for careful reflection on the limits within which the legal system may legitimately rely on such instruments. The research proceeds from the assumption that the interaction between artificial intelligence and criminal procedure is now inevitable and already underway. Rather than questioning the mere appropriateness of its use, the study seeks to identify conditions, safeguards, and regulatory models capable of integrating new technologies into the procedural system in a manner consistent with the epistemological principles of criminal proceedings and the protection of fundamental rights. From this perspective, the paper advocates the emergence of a new phase in criminal procedure, grounded in human–machine collaboration, capable of reconciling technological progress, the reliability of tools, and the safeguarding of the principles of a fair trial, within a clear and precise regulatory framework.| File | Dimensione | Formato | |
|---|---|---|---|
|
Amidani tesi corretta.pdf
embargo fino al 26/03/2028
Licenza:
Tutti i diritti riservati
Dimensione
4.69 MB
Formato
Adobe PDF
|
4.69 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/381747
URN:NBN:IT:UNIBS-381747