Springtime for EU and AI : The European commission’s Artificial intelligence act

This master thesis aims to explore and discuss a central problem within archival science, namely, how to secure transparency in AI-systems. This is done through an analysis of the European commission´s Artificial intelligence act, a proposal meant to address the ethical issues concerning AI and secure a competitive European market. Using the theoretical framework “Frames”, this paper argues that the commission perceives the Artificial intelligence act as a social response to AI as a phenomenon and challenge. This is supported with a comparison with other publications concerning artificial intelligence by the European commission. The paper also investigates the role of archives and recordkeeping in securing transparency of AI. According to the commission’s proposal, its main task is to explain AI decision or by reporting the systems design and capacity. In order to discuss possible consequences to governmental agencies, the AI act is compared with current Swedish archival law and practice. Furthermore, the referrals by the Swedish government and agencies are also included to put this discussion in a wider perspective. Although positive to the initiative, the referrals show a concern of a vague definition of AI, expanding administration and its possible obstructive effect on AI development. I argue that the concerns should be taken seriously. Instead of mindlessly inserting documentation requirements, archivist should take a step back and ask themselves what information is relevant to preserve about artificial intelligence. A difficult question to answer indeed, but necessary because the changes and risks the technology entails requires some sort of documentation in order to guarantee transparency. Therefore, archivists should explore and discuss this important and exciting topic, in order to face a revolutionary technology and future. This is a two years master's thesis in Archival science