In recent decades, digital and virtual technology has been profoundly influencing the way we live our daily, family and professional lives, becoming a sometimes invasive tool that has changed – perhaps forever – our habits and ways of communicating and living.
In the recent past, therefore, we have witnessed the implementation of new virtual environments (e.g., the Metaverse), which are posing various problems and questions in the legal sphere, given the absence of current legislation that can provide the tools to settle the disputes that are arising, with increasing frequency, in such digital environments.
Specifically, with regard to generative AI, we have witnessed a constant evolution of this technological tool since 1966, when a so-called chatbot (ELIZA) capable of interacting with human beings was developed at MIT.
The issues raised by the latest innovations
In the last two years, we are witnessing the emergence of new generative AI systems, such as Chatbot (1) and DALL-E (2), which, besides arousing great curiosity, are generating some perplexities, both from an ethical-philosophical and legal point of view.
The above examples are part of the concept of generative AI, whereby machine learning processes a huge amount of data, information and images, by means of training algorithms, in order to create new data or create new versions of existing data. Let's think, for instance, of what happens when your mobile phone or Google suggests what the rest of the word/phrase you are composing might be: this is a low-level version of generative AI.
It tends to be the case that the AI system acts upon human input and according to set algorithms. However, it should be noted that, through increasingly advanced mechanisms, AI has started to autonomously perform some typical human actions such as composing melodies, dialoguing and creating works of art.
It is precisely such artificially generated creativity that raises several questions in the field of copyright law, contemplating hypotheses in which there are entities, apart from humans, capable of producing a creative work.
The current legal framework
Is legal protection of works created by AI therefore possible? First of all, it's essential to distinguish works created by AI as a result of human input from works created independently by AI.
Although in the first hypothesis it is easy to recognize a human copyright, in the second case several difficulties arise in identifying the owner of the work and, consequently, its entitlement to exercise IP rights over it in court, given the absence of legal capacity in the AI system.
Over the years, therefore, national and European jurisprudence (UE Directive n. 2019/790, on copyright and related rights in the digital single market) has explicitly ruled out copyright protection for works created by AI, given the absence of a creative character in them, which is instead recognizable in works created by human beings.
The Italian doctrine, at the same time, has attempted to put forward some potential solutions in order to fill and regulate this gap in the law, offering some possible solutions:
- attributing ownership of the work created by AI to the person who programmed the software and/or the algorithms that provided the tools to create the work;
- considering productions created by an AI system as 'orphan' works, subject to copyright protection, but whose right holders are unknown or untraceable;
- applying by analogy what is provided for the contents of databases under Article 102-bis, Law No. 633/41 (Italian Copyright Law).
Some recent case studies
Endorsing the aforementioned doctrinal orientations, the Court of Cassation recently pronounced its opinion in the Order of the Court No. 1107 of 16 January 2023, on the subject of copyright of artificial intelligence.
Briefly, the facts of the case concerned, among others, Rai – Radiotelevisione Italiana S.p.A. (the Italian public broadcasting company), defendant for copyright infringement, concerning the graphic work ‘The Scent of the Night’, used as a set design for the 2016 Sanremo Music Festival.
Although the judges took up the constant Italian and European guidelines, they carried out an interesting in-depth study on the ‘measurement’ of the human creative contribution, during the generative process, carried out by an artificial intelligence system.
This verification, aimed at defining the extent to which the AI software has absorbed and processed the artist's creative inputs (within a structured and complex creative process), becomes extremely important for the purpose of the possible attribution of copyright protection.
The current European guidelines would also seem to be taken over by the US legal system. In February, the United States Copyright Office (USCO) issued an important copyright decision concerning certain images created by an AI system, denying registration of such works, such as the graphic novel ‘Zarya of the Dawn’ by Kristina Kashtanova.
Specifically, the examiner, while acknowledging that the applicant was the material author of the text of the work, on the other hand ruled that the individual images in the work itself were created by the non-predictive AI system Midjourney and were therefore not protectable under the US Copyright Act. This software, in fact, automatically generates results and, therefore, the inputs set by the author were considered by the examiner as mere suggestions and not instructions given to the AI system.
The need to update IP regulations
From what emerges, to date, both civil law and common law jurisdictions have in recent years appeared sensitive to this technological development and to the critical issues created by it, setting up ad hoc study commissions and issuing opinions or pronouncements that can provide principles of law applicable, by analogy, in those grey areas that still lack specific regulation.
At the European level, we recall, among others, the ‘Report with recommendations to the Commission on Civil Law Rules on Robotics’ (published on 27 January 2017 by the EU Legal Affairs Commission) and the ‘White Paper on Artificial Intelligence’ (published on 19 February 2020).
The legal framework, in any case, needs to be uniformly updated at European level in the field of intellectual property, given the absolute absence in the acquis communitaire of specific regulatory provisions applicable to the field of artificial intelligence, through the ad hoc creation of new legal principles and/or through the transposition of pre-existing ones into the digital world.
- Acronym for Generative Pretrained Transformer: a human language processing system that, through the use of advanced machine-learning algorithms, is able to generate responses that are entirely similar to human responses.
- Artificial intelligence algorithm capable of generating images from textual descriptions/queries.