Servizio sanitario nazionale, intelligenza artificiale e responsabilità medica

In recent years, Artificial intelligence (AI)-based technologies have found fertile ground in the healthcare sector. The development of effective classification and predictive algorithms, coupled with the growing digitisation of healthcare data, undoubtedly facilitates the use of these tools for various clinical tasks, particularly during the diagnostic phase. However, alongside the potential benefits of these devices—which are designed to assist healthcare professionals in their patient care duties—a number of new risks are emerging, related to the quality of the processed data, the presence of discriminatory biases, possible errors in the results, as well as the interaction between humans and AI. Specific preventive technical standards have therefore been developed and are now set out in EU Regulation 2024/1689 (the AI Act), addressing the need to ensure safe and reliable AI. In the medical device sector, these standards are supplemented by Regulation 2017/745. If these precautionary measures fail and a patient is harmed as a result of using an AI system, it must be determined which parties involved in designing and applying the device are liable. The harm may be due to human error, software malfunction, or a combination of both. The potential liable parties include the physician, who is responsible for critically monitoring the device’s operation; the healthcare facility to which the physician belongs, which bears vicarious liability for the professional’s conduct and must adopt appropriate technical and organisational measures to manage safe technologies and staff training; and finally, the manufacturer, which is liable for any defect detected in the AI system. The relationship between the healthcare professional and the AI-enabled device is particularly critical. The opacity of the machine’s decision-making process (“black box”) prevents the physician from understanding which variables the system took into account in making the recommended diagnosis. Therefore, the development of explainable AI systems is crucial to enable effective oversight of the device’s operation. In terms of the healthcare facility, integrating AI-enabled devices into clinical practice involves meeting a series of preparatory organizational obligations, in accordance with Articles 26 and 27 of the AI Act. Failure to comply with these provisions may result in the entity being held liable under Article 1218 of the Italian Civil Code if organisational failure causes harm to the patient. After all, it is precisely the facility that has the concrete ability to mitigate the risks arising from the use of a given system by intervening in the development, implementation, and supervision of AI. To this purpose, the Spanish model of strict liability of healthcare facilities for damages caused by a "defective" medical service has been analysed from a comparative perspective pursuant to art. 148 of the Texto Refundido de la Ley de Consumidores. Lastly, the manufacturer’s role is relegated to the background in this context due to the presence of other parties that interact directly with the AI system and with the injured patient. However, when the harm can be attributed, at least partially, to software malfunction or a lack of adequate instructions and training, it is possible to raise a claim for concurrent liability for a defective product, in accordance with EU Directive 2024/2853.