The European AI Act: Generative AI and Foundation Models

Beskrivning

Stefan Larsson presents his study on how generative AI and so-called foundation models are handled in the process of developing the European AI Act. As part of a study on the pacing problem between regulation and innovation in the area of AI, Stefan presented a mapping of the different terminologies around the AI definition found in the different version of the AI Act from April 2021 to June 2023, with a particular focus on aspects of generative AI, general purpose AI and foundation models.

The conclusions in the presentation include the fact that some of the concepts are novel and with some ambiguity, can also seemingly be partly overlapping, and without a clear consensus in how industry would perceive them. Furthermore, the general purpose AI, introduced after the Commission's version, may be hard to fit into the risk levels introduced in the Commission's version. Transparency is presented as a key idea for regulation of foundational models, but is also likely to bring some unintended consequences. Overall, the AI Act proposals show signs of the pacing problem in terms of its mix of fixation and flexibility, via delegated acts, implementation acts, sandboxes, standards to be developed.

Period	2023 sep. 26
Evenemangstitel	Regulating High Risk AI in the EU: AI Lund Fika to Fika Workshop
Typ av evenemang	Workshop
Plats	lund, SverigeVisa på karta
Omfattning	Internationell