Justice: disclosure requirement for damages suffered by the AI
The injured party is entitled to access the technical documents. The right to bring a direct claim against the insurance company has been recognised
Key points
A tailored procedure for damage caused by artificial intelligence, whether contractual or non-contractual. The draft legislative decree approved by the Council of Ministers on Wednesday addresses not only criminal matters (investigations, offences, and biometric identification procedures for fugitives and suspects), but also in the civil sphere with a package of measures dedicated to the unprecedented form of liability for failure to exercise due care regarding the use of artificial intelligence systems.
Access to documents
The Government explains that the measures relating to civil protection aim to redress the balance for those who suffer harm as a result of an AI system, by overcoming technological opacity and information asymmetries without imposing any new substantive obligations on businesses. The decree aims to strengthen access to justice for the injured party in a technically complex field, where reconstructing how the system works and establishing a causal link can prove particularly difficult.
The following aspects are therefore recognised: access to the technical documentation of the system, to enable the injured party to understand the key features of the AI used, and the presumption of causality, which lightens the burden of proof without eliminating it entirely.
Evidence
Specifically, upon application by the party claiming to have suffered the damage, the court shall order the other party or any third party in possession of such evidence to produce evidence regarding the functioning of the artificial intelligence system, where the request sets out facts and elements such as to render its validity plausible, including with regard to the link between the result produced by the artificial intelligence system and the damage alleged.
The evidence includes a series of documents required under the European Regulation on Artificial Intelligence: records and documentation relating to the risk management system, as well as information on the parameters and procedures for human oversight.


