Outputs generated by artificial intelligence systems: ownership and liability under the current legal framework

The increasing use of artificial intelligence systems in the production of digital content — including texts, images, software and predictive models — calls for an increasingly careful assessment of the legal regime applicable to the outputs generated by such systems.

Under the current copyright framework, protection presupposes the existence of a human creative contribution. Article 1 of Law No. 633 of 22 April 1941 protects intellectual works that have creative character and are attributable to an author. This requirement gives rise to clear difficulties in the case of content autonomously generated by artificial intelligence systems, where it may be complex to identify a genuinely relevant human creative contribution.

In such cases, significant questions arise concerning the ownership of economic rights over the outputs produced. The absence of an identifiable human author and the difficulty of bringing algorithmic processing within the traditional concept of creativity may, in some cases, lead to the exclusion of copyright protection under the classic criteria of copyright law. This scenario creates an area of uncertainty that is particularly relevant for economic activities based on the production of digital content or the use of generative systems.

The critical issues are particularly evident at contractual level. In software development agreements, SaaS agreements and, more generally, in relationships for the provision of technology services based on artificial intelligence systems, it becomes essential to expressly regulate the ownership of the outputs generated by the system.

A generic clause granting the client “ownership” of the results of the processing may prove insufficient if it does not clarify the regime applicable to the datasets used for training, the generative models and any content derived from pre-existing works.

A further critical aspect concerns liability for possible infringements of intellectual property rights. Where the output generated by an artificial intelligence system reproduces, even only in part, protected works or pre-existing content, the issue arises of identifying the party responsible.

In the absence of specific legislation, the assessment must be carried out on a case-by-case basis in light of the general principles of contractual and non-contractual liability, with particular attention to the risk allocation clauses contained in agreements between developers, service providers and users of the systems.

The growing spread of artificial intelligence technologies therefore requires a particularly prudent approach to the structuring of contractual relationships and the management of intellectual property rights. In this context, the absence of a fully defined regulatory framework increases the role of contractual provisions in governing the ownership of outputs and allocating the risks arising from the use of generative systems.