I requisiti per l’addestramento degli strumenti di Intelligenza Artificiale e il deposito legale delle risorse digitali: una riflessione sul contesto normativo italiano e il ruolo delle biblioteche

In 2022, the National Central Library of Florence launched an experimental project to train the Annif tool for automatic semantic indexing in the Italian language, developed and maintained by the National Library of Finland. The project revealed a more general lack of the prerequisites in Italian libraries for using not only the more traditional Machine Learning tools but also new Artificial Intelligence systems. The causes can be traced, on the one hand, to deficient legislation on the legal deposit of digital resources and, on the other hand, to the fragmentation and multiplication of vocabularies and thesauri, which respond to different ontologies and are structured in formats not always suitable for web interoperability. The purpose of this contribution is to initiate a reflection on these issues, attempting to outline the role that libraries could play in the ethical and conscious development and use of Artificial Intelligence systems, with an in-depth look at the ongoing project called V.I.CO. - Vocabulary of Controlled Identities.