A new Wolters Kluwer commentary argues that the international three-step test—requiring exceptions to be limited to special cases that don't conflict with normal exploitation or unreasonably prejudice rightsholders—should be central to how courts and regulators assess AI training data practices. The piece frames market dilution, not just verbatim copying, as the underlying economic harm the test is meant to catch, a framing that could sharpen fair-use and TDM-exception debates on both sides of the Atlantic.
For data licensing markets, this suggests scrutiny may increasingly turn on aggregate market effects rather than case-by-case copying analysis.
Training Data, Market Dilution, and the Elephant in the Room: Why the Three-Step Test Matters for Generative AI
