As is well known, the use of artificial intelligence (AI) in content creation has marked a momentous turning point in recent years, so much so that it is now capable of generating and producing works of different kinds, including musical compositions, literary texts, visual artworks and design projects.
For example, thanks to the use of artificial intelligence in the music industry, starting from musical bases generated by algorithms and refined by musicians, it has been possible to compose songs of various genres, from classical music to modern pop. Such AI platforms can be used for free by anyone who wants to generate new musical compositions; in fact, it is enough to download the appropriate program from any search engine and briefly describe the content you want to listen to for the desired musical text to be played.
Not only that. AI is also being particularly used in the field of visual art and in fact has made it possible to recreate works of art exhibited and sold at auctions as one-of-a-kind pieces, among which it is worth mentioning some collections made with the help of AI exhibited in the well-known MoMA (Museum of Modern Art) museum in New York.
And even in the field of literature, artificial intelligence is being used to create poems, short stories and entire novels simply by asking the “machine” to analyze and study existing literary collections.
Such artificial intelligence systems are better known as “Generative Artificial Intelligence,” as the content is created and generated by using complex algorithms and machine learning models. These are systems designed to learn from large amounts of existing data through which they are then able to generate new content.
It goes without saying that this extraordinary ability of AI astonishes but also worries, as it challenges the traditional notion of authorship and raises a number of questions about the ownership of rights over such creations and works.
Indeed, one of the main difficulties related to this content creation process, is whether the work generated by generative AI can be considered “original” according to the traditional notion of copyright and whether the AI creation process represents a “reworking / transformation,” legitimate or not, of the original work.
Since the algorithm uses third-party, copyright-protected material, it could be argued that generative AI is capable of generating derivative work, which, as is well known, is a new creation that incorporates or modifies a pre-existing work already protected by law. In this context, if a generative AI algorithm is trained, for example, with copyrighted pieces of music, it is open to question whether the result obtained violates the copyrights of the original works to the extent that it constitutes an unauthorized reworking of them.
If, again for the sake of argument, one was to classify the work created by generative AI as an autonomous and original work, one would then have to ask which party would be entitled to the relevant intellectual property rights over that creation.
However, the current copyright rules are unable to provide any guidance on such questions, given that they were conceived at a time when it was hardly conceivable that any content could be created by anyone other than a human being.
As a result, identifying the artificial intelligence system as the author of the work, and thus the holder of the relevant copyright, would raise a number of complex legal issues, certainly including that of legal capacity, which, as is well known, is the ability of a subject to be the holder of subjective legal situations and thus of rights and duties. This requirement is inevitably lacking when discussing artificial intelligence systems.
For the above-mentioned reasons, the question therefore arose as to whether the authorship of the work should therefore be attributed to the developer of the generative artificial intelligence system.
However, this statement poses a further problem: the creativity and originality of the work is not directly ascribable to the action of the developer, rather than to the artificial intelligence system itself, or in any case it is particularly difficult to identify the creative contribution of one or the other.
The same also applies to the owner of the data used in the training/learning phase of the artificial system; in fact, even if one wanted to identify this subject as the owner of the copyright, the dataset provided by him, in the absence of a learning model, would in any case not be able to generate the work and therefore once again there would be a risk of attributing and recognizing rights that are not exclusively due to a human being.
Case law has attempted to provide an answer to these numerous questions. Indeed, an initial decision by the Copyright Office in the United States of America denied any copyright recognition to works created entirely by a generative AI system, clearly stating that for a work to be protected by copyright, there must be a human author involved in its creation.
This orientation has also been confirmed by more recent court rulings in the US and, in particular, by a federal court in Washington D.C. in 2023.
In Europe, a similar conclusion has also been reached: the Supreme Court recently ruled on this issue in Judgment no. 16346 of 2023, regarding the definition of authorship for works created by means of AI systems and the level of integration of the creativity of the artist in the use of the software. In the opinion of the judges of the Court, it is necessary to assess on a case-by-case basis whether, during the generative process of the work, the software in use has integrated with the creativity of the human user.
The difference between a simple automatic generation of content by generative AI systems and a work that is an expression of the creativity of the artist through digital AI is therefore of fundamental importance. Only in the latter case, by demonstrating therefore that there has been a significant human creative contribution, is it possible to find ways of enforceable protection under copyright law.
In this context, the need for an intervention by the legislator to redefine key concepts and prepare new copyright regulations for the AI phenomenon, which is destined to grow exponentially, is unquestionable.
One solution that is worthy of attention could be the drafting of new forms of licenses and copyrights that would better suit today’s reality and represent a meeting point between technological innovations and copyright protection. For instance, we could consider introducing specific licenses for AI-generated works or developing new ways of enforcement in which the collaboration between human ingenuity and the aid of AI is clearly recognised.
Finally, more cooperation at European and international level is desirable, considering the global dimension of technological innovation and distribution of creative works. Digital technologies, such as the internet and content distribution platforms, are by their very nature global and therefore require a coordinated response from authorities in different countries.
It will be crucial to work together to establish standards and rules that work globally, while ensuring that the copyrights of different jurisdictions are respected.