The research project focus on the impact that scientific and technological developments — particularly in the fields of robotics and artificial intelligence — have had, and continue to have, on contractual relationships and the system of civil liability at the European level, with specific regard to the healthcare sector. Contemporary computational power, in light of the ability to process massive volumes of data, enables the analysis of clinical data on a scale that surpasses and expands beyond the experience of the individual healthcare professional; this highlights the benefits arising from the application of such systems in terms of diagnostics, prevention, surgery, automation of healthcare processes, as well as increased efficiency and the reduction of human error. At the same time, however, the increasingly extensive applications of robotics and artificial intelligence in the healthcare field give rise to continuous legal — as well as ethical and social — questions, which require legal scholars and practitioners constantly to identify appropriate remedies capable of filling the regulatory gaps created by the rapid evolution of intelligent systems, an evolution that may profoundly affect fundamental rights such as the right to health, freedom, dignity, and self-determination. Among the emerging issues, two characteristic profiles inevitably intersect. On the one hand, there is the problem of the so-called black box, namely the intrinsic opacity of AI systems, whereby only the input and the output can be known without any possibility of understanding the logical mechanisms followed by the algorithm; consequently, it becomes impossible to determine where and how the error or malfunction that caused the damage may have arisen. On the other hand, one must take into account the long chain of actors involved in the production, installation, training, and use of AI systems. In the field under consideration, healthcare professionals operate within a network of individuals, institutions, and technological tools that cannot be entirely mastered by them and upon which the outcome of the therapeutic treatment administered to the patient depends; consequently, the allocation of damage moves away from the traditional system of attributing liability based on the foreseeable consequences of one’s own conduct. The coexistence of these factors makes it particularly complex to identify an appropriate liability regime for AI systems. Although the AI Act introduces a pyramid-shaped risk-based model regulating the permitted and prohibited uses of such systems, as well as obligations concerning conformity assessment, risk management, and other duties imposed upon providers and deployers, the issue of identifying the appropriate liability standard remains unresolved and is therefore still governed by the various national legal systems. One proposed solution is represented by the new Product Liability Directive — which was initially intended to enter into force alongside the AI Liability Directive, the latter having been withdrawn in February of this year — expressly extending the notion of “product” to software and AI systems, thereby broadening the range of potentially liable subjects throughout the digital supply chain and offering a viable solution where the defect lies within the algorithm itself. However, not every instance of damage can be linked to a product defect; therefore, situations arise that require an analogical application of Articles 2049, 2050, and 2051 of the Italian Civil Code, although such an approach does not appear to be sufficient. With regard to these issues, Delegated Law No. 132 of 2025 itself draws attention to the applications of AI across various sectors, including healthcare (Article 7), and to the obligations imposed on deployers in order to minimise risks and improve patient health. It nevertheless remains necessary to develop ad hoc legislation grounded in both European and national law, together with a necessary reconsideration and adaptation of the Gelli-Bianco Law.
La ricerca nonché l’elaborato si concentrano sull’impatto che i progressi scientifici e tecnologici— in particolare della robotica e dell’intelligenza artificiale — hanno avuto e continuano ad avere nei rapporti contrattuali e nel sistema della responsabilità civile a livello europeo, con specifico riguardo al settore sanitario. La potenza di calcolo odierna, alla luce della capacità di processare volumi massivi di dati, consente l’analisi dei dati clinici contemporanei, superando e ampliando l’esperienza del singolo operatore sanitario; ciò evidenzia i benefici dell’applicazioni di tali sistemi, sotto il profilo della diagnostica, della prevenzione, della chirurgia, dell’automazione dei processi di cura e per quanto concerne l’aumento dell’efficienza e la riduzione dell’errore umano. Allo stesso tempo, tuttavia, le più vaste declinazioni della robotica e dell’intelligenza artificiale in ambito sanitario comportano l’insorgenza di continui interrogativi giuridici (ma anche etici e sociali), che impongono all’interprete di trovare costantemente rimedi opportuni per colmare le lacune normative derivanti dalla rapida evoluzione dei sistemi intelligenti che può incidere profondamente sui diritti fondamentali quali il diritto alla salute, alla libertà, alla dignità e all’autodeterminazione. Fra le problematiche emergenti, si intersecano inevitabilmente due profili caratteristici: da una parte, il problema della cosiddetta black box - ovvero l’opacità intrinseca dei sistemi di AI che permette di conoscere unicamente l’input e l’output senza poter comprendere i meccanismi logici percorsi dall’algoritmo- per la quale conseguentemente non risulta possibile individuare dove e come si sia eventualmente generato l’errore o il malfunzionamento che ha cagionato il danno; dall’altra, si deve fare i conti con la lunga catena dei soggetti coinvolti nella produzione, installazione, addestramento e utilizzo dei sistemi di AI. Per quanto attiene all’ambito in oggetto, il professionista sanitario si trova immerso in una rete di persone, strutture e tools che non possono da lui essere padroneggiati interamente e dai quali dipende l'esito del trattamento terapeutico somministrato al paziente; di conseguenza, l'allocazione del danno si allontana dal classico sistema di imputazione delle conseguenze prevedibili del proprio agire. La coesistenza di questi fattori rende complessa l’individuazione di un regime di responsabilità adeguato per i sistemi di AI. Per quanto l’AI Act introduca un modello piramidale basato sul rischio che disciplina gli utilizzi consentiti e vietati di tali sistemi, nonché gli obblighi riguardanti la valutazione di conformità, la gestione del rischio (e altri ancora) in capo ai produttori e agli utilizzatori, allo stato dell’arte, rimane aperto il problema della qualificazione dell’idoneo criterio di responsabilità che pertanto viene disciplinato dalle singole normative nazionali. Una soluzione proposta è rappresentata dalla nuova direttiva sulla responsabilità per danno da prodotti difettosi – che inizialmente doveva entrare in vigore contestualmente alla direttiva sulla responsabilità da intelligenza artificiale, a febbraio di quest’anno ritirata- che estende espressamente la nozione di prodotto ai software e ai sistemi di AI, ampliando così i soggetti potenzialmente responsabili lungo la filiera digitale, offrendo una soluzione percorribile quando il difetto risieda nell’algoritmo. Non ogni danno però può essere correlato al difetto del prodotto, pertanto, si prospettano fattispecie che richiedono un’applicazione analogica degli artt. 2049, 2050, 2051 c.c. che tuttavia non appare sufficiente. Riguardo a tali problematiche, la stessa legge delega n. 132 del 2025 pone l’attenzione sulle applicazioni dell’AI nei vari settori, tra cui quello sanitario (art. 7) e su oneri in capo agli utilizzatori al fine di minimizzare il rischio, migliorando la salute dei pazienti. Risulta tuttora necessaria la definizione di una normativa ad hoc che ponga le basi sulla normativa europea e nazionale con una necessaria rielaborazione della legge Gelli-Bianco.
ARTIFICIAL INTELLIGENCE SYSTEMS, HEALTHCARE SERVICES AND CIVIL LIABILITY
GIGOLA, MONICA
2026-06-16
Abstract
The research project focus on the impact that scientific and technological developments — particularly in the fields of robotics and artificial intelligence — have had, and continue to have, on contractual relationships and the system of civil liability at the European level, with specific regard to the healthcare sector. Contemporary computational power, in light of the ability to process massive volumes of data, enables the analysis of clinical data on a scale that surpasses and expands beyond the experience of the individual healthcare professional; this highlights the benefits arising from the application of such systems in terms of diagnostics, prevention, surgery, automation of healthcare processes, as well as increased efficiency and the reduction of human error. At the same time, however, the increasingly extensive applications of robotics and artificial intelligence in the healthcare field give rise to continuous legal — as well as ethical and social — questions, which require legal scholars and practitioners constantly to identify appropriate remedies capable of filling the regulatory gaps created by the rapid evolution of intelligent systems, an evolution that may profoundly affect fundamental rights such as the right to health, freedom, dignity, and self-determination. Among the emerging issues, two characteristic profiles inevitably intersect. On the one hand, there is the problem of the so-called black box, namely the intrinsic opacity of AI systems, whereby only the input and the output can be known without any possibility of understanding the logical mechanisms followed by the algorithm; consequently, it becomes impossible to determine where and how the error or malfunction that caused the damage may have arisen. On the other hand, one must take into account the long chain of actors involved in the production, installation, training, and use of AI systems. In the field under consideration, healthcare professionals operate within a network of individuals, institutions, and technological tools that cannot be entirely mastered by them and upon which the outcome of the therapeutic treatment administered to the patient depends; consequently, the allocation of damage moves away from the traditional system of attributing liability based on the foreseeable consequences of one’s own conduct. The coexistence of these factors makes it particularly complex to identify an appropriate liability regime for AI systems. Although the AI Act introduces a pyramid-shaped risk-based model regulating the permitted and prohibited uses of such systems, as well as obligations concerning conformity assessment, risk management, and other duties imposed upon providers and deployers, the issue of identifying the appropriate liability standard remains unresolved and is therefore still governed by the various national legal systems. One proposed solution is represented by the new Product Liability Directive — which was initially intended to enter into force alongside the AI Liability Directive, the latter having been withdrawn in February of this year — expressly extending the notion of “product” to software and AI systems, thereby broadening the range of potentially liable subjects throughout the digital supply chain and offering a viable solution where the defect lies within the algorithm itself. However, not every instance of damage can be linked to a product defect; therefore, situations arise that require an analogical application of Articles 2049, 2050, and 2051 of the Italian Civil Code, although such an approach does not appear to be sufficient. With regard to these issues, Delegated Law No. 132 of 2025 itself draws attention to the applications of AI across various sectors, including healthcare (Article 7), and to the obligations imposed on deployers in order to minimise risks and improve patient health. It nevertheless remains necessary to develop ad hoc legislation grounded in both European and national law, together with a necessary reconsideration and adaptation of the Gelli-Bianco Law.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.



