The protection of the output generated by AI systems
The extraordinary deployment of the generative AI systems raises several legal issues, including the twofold problem of the protectability of the output produced by AI systems and the liability in the event that these works infringe the rights of third parties.
These issues challenge our society with essential ethical questions before even considering the legal ones. And the choices that will be made by national legislators in this area will profoundly affect the economic and social framework of the next few decades.
First, it should be noted that the output generated by AI systems can take the most disparate forms: images can be created from simple text descriptions, but also sounds, videos or graphics can be produced just from data sets. Moreover, such outputs are frequently the result of an autonomous work created by AI, without any human participation. Recently, works generated by AI systems have even received prestigious awards: for instance, the short film The Crow won the Jury Prize at the Cannes Short Film Festival and, according to a report in the New York Times, an AI-generated artwork won the Colorado State Fair’s annual art competition.
Copyright ownership of works made by Al: Italian and non-EU legislation
However, from a legal standpoint, the issue is much more intricate than just awarding a prize. Given this complexity, we can briefly point out that to this day – with a few exceptions – most national legislations, including non-European ones, tend not to recognize copyright ownership of intellectual works created by machines.
The only jurisdictions that expressly provide protection for computer-generated works are Hong Kong, India, Ireland, New Zealand, South Africa and the UK. In particular, in the UK, the Copyright Designs and Patent Act 1988 provides that copyright of a work created by an AI machine is vested in the person who has organized the machine’s functions in such a way as to enable the machine to generate the work.
One of the main arguments in favor of recognizing protection for content generated by AI systems is the investment of economic resources, and other types of resources in the development and training of the algorithms that underlie such systems. In other words, granting protection to AI-generated content produced would also indirectly protect the significant investments made by market players.
Some commentators have, however, observed that such an approach would be more akin to the recognition of related rights, such as those of phonogram producers, instead of a proper copyright protection. On the other hand, to have protection under copyright law, many national legislations require human contribution (in Italy, Law no. 633/1941 – the Copyright Law – expressly refers to “intellectual work”; the US copyright law is limited to the protection of the “author’s original intellectual conceptions”, and so on). However, since these legal frameworks were devised before the development of today’s sophisticated AI systems, there are many grey areas and difficulties related to their interpretation, which are likely to lead to possible litigation.
The US and Italy: Two conflicting approaches on copyright protection of AI generated works
In the first months of 2023, two cases, one in the US and the other one in Italy, provided conflicting approaches on the topic of copyright protection of works created using AI tools.
On one hand, the US Copyright Office cancelled a certificate of registration previously granted to the graphic novel Zarya of the Dawn on the grounds that the author of the comic strip, Kristina Kashtanova, had used well-known image-generating software to create the sequence of drawings that made up the comic strip.
The communication sent by the US Copyright Office clarified that Kristina Kashtanova “author of the work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements. That authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the […] technology are not the product of human authorship”. Under US law only certain elements of the work are susceptible to copyright protection, which covers just “the fruits of intellectual labor”. Therefore, the Office cancelled the original registration certificate and issued a new one covering only the material effectively created by the author.
In Italy, a recent decision of the Italian Supreme Court dealt, albeit incidentally, with the issue of the protectability of a work created with the aid of software.
The dispute arose from RAI’s use of the image of a digital flower, which was retrieved online, as the centerpiece of the set design of the Italian Music Festival (Sanremo) in 2016, without the author’s authorization. The broadcaster brought proceedings before the Italian Supreme Court claiming that the second-instance decision of the Court of Appeal had erroneously qualified as intellectual work an image generated by a software and not attributable to a creative idea of its supposed author.
Contrary to the conclusions of the US Copyright Office, the Italian Supreme Court clarified that the use of software is compatible with the processing of a work of art, specifying that in such cases the creativity element should be rigorously scrutinized to verify on a caseby – case basis the extent to which the use of the tool absorbs or replaces the creative elaboration of the artist. In other words, the Italian Supreme Court does not exclude that works generated through software or other computer mechanisms may be awarded copyright protection. Therefore, if the AI system is considered a mere tool through which a human subject creates works, according to this approach, the content could still be considered an expression of the intellectual creativity and personality of its author.
Plagiarism and infringement: The protectability of AI-generated works inspired by or derived from pre-existing works
In light of the above, another aspect to be evaluated is the possibility that an original work created by AI could be considered the result of plagiarism or infringement of another pre-existing work.
A similar issue has already been brought before the US courts. In particular, in January 2023, the artists Sarah Andersen, Kelly McKernan and Karla Ortiz filed a class action aimed at challenging the legitimacy of a certain image generator software. According to their claim, these AI systems would have drawn on billions of images and photographs, including their own, for their “training”, without first obtaining the consent of the legitimate owners.
The artists’ claims of unlawfulness are not limited to the AI learning modalities, but also extend to the outputs generated by them, which reflect the style and characteristics of their works. On this note, the question of protectability for works inspired by or derived from others and their relationship with the original work becomes relevant.
This issue is certainly not new and – especially in the field of art – can present significant margins of uncertainty.
The Italian legal system covers this type of works under Art. 4 of the Italian Copyright Law, which protects elaborations of works without affecting the rights of the original work. Under Art. 7 para. 2 of the same law, the author of the elaborated work is considered to be the author within the limits of their own work. Furthermore, art. 18 of the Italian Copyright Law provides for the author’s exclusive elaboration right, which includes all forms of modification, development and transformation of the work provided for under Art. 4, implying the need to obtain the artist’s consent for any subsequent creative elaboration.
It follows that a derivative work enjoys independent protection under copyright as a creative elaboration, provided that it has been authorized by the author of the original work. On the contrary, no authorization is required if the new work is a parody, which is to be understood in a broad sense. Italian case-law has clarified that parodic, humorous or ironic works, and in general works that revisit another person’s work, are parodic to the extent that they change the meaning of the original work, so as to rise to the role of independent work of art and as such worthy of autonomous protection. To achieve this result, it is therefore necessary to consider the derivative work as a whole and assess whether, although inspired by the original, it diverges from it to convey a different message.
When a transforming activity is carried out by AI systems, the question becomes even more complicated. On one hand, the issue of copyright protection of an AI-generated work becomes once more relevant, which then raises the question on whether the activity of a machine can effectively qualify as “creative elaboration”, especially when any human contribution is missing. On the other hand, as a general principle, it would be difficult to imagine a parody work realized by an AI system, since one would have to admit that they are capable – on their own – of changing the meaning of a pre-existing work. If this was the case, we would then be faced with AI systems equipped with critical-thinking ability, which could be defined for all intents and purposes as sentient.
