The U.S. Copyright Office rejects a single answer for AI training and fair use
The U.S. Copyright Office released the pre-publication Part 3 of its AI report on generative-model training. It treated fair use as fact-specific, highlighted the importance of source acquisition and market harm, supported voluntary licensing markets and did not recommend a new blanket compulsory licence at that stage.
Why it matters: Training-data risk depends on how works were obtained, what the model does with them and whether outputs substitute for protected markets—not merely on the label “AI training.”