A Position Paper on AI and Copyrights in Cultural Heritage and Research (EU and UK)

The long-term, highly divisive discussion between the entertainment industry and the AI sector around the use of copyrighted material for the training of machine learning applications has gained traction with the recent advent of generative AI systems. However, in this debate, the voice of cultural heritage institutions and research libraries is hardly considered. Major publishing houses have begun to reserve the right to perform text and data mining (TDM) on digital assets licensed by them, which is relevant for the development of machine learning applications. Such reservations impair the access of cultural heritage institutions (CHIs) and research libraries’ patrons, such as start-ups, small and medium-sized enterprises and companies belonging to the cultural sector. In this discussion paper, we propose a set of differentiated solutions which potentially open up ways in which digital assets which are under copyright may be used for TDM. These solutions consist in the development of new licences, the use of a technical protocol, and in the provision of infrastructures which allow for performing TDM on copyrighted digital assets differentiated according to data providers, data consumers and to the tasks for which machine learning applications are being trained. These proposals aim at furthering the debate on open data – keeping them as open as possible while respecting the interests of right holders and data consumers alike.