29 Mar Copyright and generative artificial intelligence

Authors: Miriam Andrea Fadda, Francesca Tugnoli, Eleonora Margherita Auletta.

The widespread application of artificial intelligence (AI) presents challenges in many areas, including in the legal field. New applications using AI are forcing legal experts to reflect on how to adapt analog legal cases  to new technologies. Such technologies are becoming increasingly advanced and are dominating the market in a situation characterized by a lack of specific regulation.

Generative artificial intelligence

Let’s start with a simple question: what is generative artificial intelligence?

Generative AI is a branch of the broader concept of “artificial intelligence”[1] that specifically enables the creation of various types of content from a synthetic data set: from newspaper articles to images, from videos to music, just to name a few examples.

When we talk about generative AI, reference is made to a variety of processes aimed at feeding a system that has been trained by specific datasets which are related to the work to be generated by the application based on specific commands (so-called prompts).

The training to which AI-based technologies are subjected differs based on the specific underlying software: generative AI, for example, requires quite different training from a machine learning-based software. Training modalities are very complex and are based on a variety of trials. One can go so far as to say that the training of AI never ends, due to the possible exploitation of new datasets to continue to expand its store of information on the basis of which it provides the required output.

Intellectual work and copyright law

As is well known, Italian Law No. 633/1942[2], better known as the Intellectual Property (or IP) Law, protects intellectual works, which, according to Article 1 of the aforementioned legislation, must be characterized by the following cumulative requirements:

• originality: the work, in order to be protected by law, must be the result of the activity of human ingenuity and must therefore reveal the author’s personality;
• objective novelty: the work must be characterized by elements that unequivocally differentiate it from any other work.

In order to protect a work created with generative AI, therefore, it is necessary to verify whether the output generated by AI-based software can be considered an extrinsic embodiment of the intellectual creation of the person who provided the prompts that, in turn, led to the result to be subject to protection.

The Italian Supreme Court’s decision

On whether or not a result created by generative artificial intelligence can be qualified as a work of authorship, the Supreme Court of Cassation recently (indirectly) expressed its opinion in its Order No. 1107 of 16 January 2023[3]. The matter under dispute concerned the alleged copyright infringement of a work used as a fixed set design for a music festival, which had been generated through the use of an AI-based software.

Despite the declared inadmissibility of the appeal in question, since it was sent after the given deadline, the Italian Supreme Court had an opportunity to express an important principle of law useful for the resolution of the question at hand. In order to ascertain whether protection of the work created through generative artificial intelligence should be granted, it is necessary to preliminarily assess whether and to what extent the use of the software has “absorbed” the creativity of the artist who had chosen the algorithm to be used. Therefore, as to the “protectability” under IP law of works generated by artificial intelligence systems, it will be necessary to distinguish between works generated through artificial intelligence systems, with respect to which human creativity assumes significant prominence, and works generated by AI systems, for which human creative input is marginal. Only in the former case, therefore, will the work be protectable under the traditional categories of copyright; moreover, both the moral right to be considered an author and the rights of economic exploitation of the work will have to be recognized to the artist as a natural person.

Conclusions

In conclusion, in the near future, in order to offer a form of protection to works created with the help of software that exploits generative artificial intelligence, it will be necessary to ascertain when the latter can be considered a work of authorship, based on the application of the above criteria.

[1] The definition of “artificial intelligence” is based on wanting to harness complex computers and machines to mimic the problem-solving ability of the human mind. As early as 1950, Alan Turing, father of computer science, in his “Computing Machines and Intelligence,” questioned whether machines were capable of thinking. This very reflection gave rise to the “Turing test,” according to which a human being asks questions and, if he or she cannot distinguish whether the answers come from a machine or another human being, then the machine that formulated them can be considered to have a form of intelligence. A more modern and precise definition is later given by John McCarthy, the U.S. computer scientist who won the Turing Prize, in 2004, who describes it as the science of creating and engineering intelligent machines and, especially, intelligent computer programs.

[2] Law No. 633 of 22 April 1941.

[3] Supreme Court, Sec. I Civ., Order No. 1107 of 16 January 2023. In this Order, the Italian Supreme Court made it clear, incidentally, that the use of digital technology for the creation of a work does not in itself preclude the possibility of qualifying the work as a work of authorship, unless the outcome of a factual assessment in which the rate of creativity has been rigorously scrutinized demonstrates that the use of technology has “absorbed” the creativity of the artist.