The United States Copyright Office (USCO) released a 108-page "pre-publication" report on May 9, 2025, addressing the unauthorised use of copyrighted materials for training generative AI systems. The report represents the final instalment of the Office's three-part series examining the intersection of artificial intelligence and copyright law, analysing the contentious fair use doctrine in the context of AI training.
The circumstances surrounding the pre-publication report's release are unusual, as President Donald Trump dismissed Dr Carla Hayden, the Librarian of Congress, one day before its publication, then fired Shira Perlmutter, the Register of Copyrights, the day after. The report establishes that using copyrighted works to train AI models may constitute prima facie infringement of several exclusive rights, particularly the right of reproduction, which is implicated during system development, data collection, curation, and training. Regarding transformativeness, the Office determines that training a generative AI foundation model on a large and diverse dataset will often be transformative, but this factor varies in degree—using copyrighted works to produce content that competes with the originals in existing markets rarely qualifies for fair use protection.
The report analyses the four statutory fair use factors, noting that guardrails implemented by AI models to prevent infringing outputs weigh more favourably in the fair use assessment. On market impact, the Office identifies three potential market effects: lost sales, market dilution, and lost licensing opportunities. The report ultimately concludes that government intervention is not necessary at this time, as the fair use doctrine and the voluntary licensing market provide sufficient frameworks to address the issues.
Sources:
1.
Roanie Levy
2.
Cynthia Cole, Adam Aft, Avi Toltzis
3.
Stuart D. Levi
