Reuters' year-in-review points to a maturing body of case law testing whether AI training on copyrighted works counts as fair use, and how piracy-sourced datasets and downstream market harm factor into that analysis. For data licensing teams and AI developers, the emerging distinctions—between transformative training and reproducing expressive content, and between lawfully acquired versus pirated corpora—are becoming the actual battle lines rather than abstract debate.
Expect the contours from these decisions to shape how companies document data provenance and structure licensing deals going forward.
Copyright Law in 2025: Courts begin to draw lines around AI training, piracy, and market harm
— Reuters
