ABSTRACT
Throughout human history, whenever a new device has been invented, society has continued to move forward, never regressing to an earlier stage. Therefore, when we are analysing the problems posed by artificial intelligence (AI) systems and, specifically, those posed by generative AI (GenAI), we must assume that there will be no regression to the pre-AI era. These systems have asserted themselves in the market for three main reasons: massive public use (at scale); development or acquisition by major market players-for example, there is currently a race for domination between Microsoft and Google-also considering non-Western countries for which we do not always have precise information (eg China or India) (platformisation); and the use of massive datasets, and therefore the use of big data, to power the machines (datafication). In the last few years, many studies have investigated the legal issues relating to copyright law and AI, with specific reference to the training of AI systems. One of the main issues arises from the possibility that machine-learning (ML) processes infringe intellectual property (IP) rights and copyright owners’ rights. In this paper, I look at GenAI models, which are trained using large amounts of data (both personal and non-personal data, such as literary and artistic works) obtained by using automated software tools (ie data-scraping technologies) to look for content freely available on the Web, even content that is potentially protected by copyright law. My main aim is to analyse the current legal framework, with reference to the AI Act, to evaluate approaches that strike a balance between the rights of the owners of the works.
Riccio, Giovanni Maria, Artificial Intelligence and Copyright: Rights and Remuneration (April 13, 2024). Forthcoming in R Potenzano and R Bieda, Legal Challenges of Disruptive Technologies, Nomos, 2024.
Leave a Reply