Open Future’s Alek Tarkowski, writing in March about Europe’s AI Act:

Overall, the AI Act does not introduce meaningful obligations for training data transparency, despite the fact that they are crucial to the socially responsible development of what the Act defines as general purpose AI systems.

Tarkowski’s post is nuanced, and well worth a read. My mind kept drifting to the scholarly-publishing case—in which scholars’ tracked behavior, citation networks, and full-text works might train propriety models built by the likes of Elsevier. As Tarkowski hints here—echoing Open Future’s July 2023 position paper—open science norms around data sharing should be brought to bear on legislation and regulation. The case for FAIR-like principles to apply to models trained on scholarly data is stronger still.