ARTICLE
4 June 2025

U.S. Copyright Office Releases Third Report On AI And Copyright Addressing Training AI Models With Copyrighted Materials

CM
Crowell & Moring LLP

Contributor

Our founders aspired to create a different kind of law firm when they launched Crowell & Moring in 1979. From those bold beginnings, our mission has been to provide our clients with the best services of any law firm in the world through a spirit of trust, respect, cooperation, collaboration, and a commitment to giving back to the communities around us.
The preliminary Report takes the position that "making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets...
United States Intellectual Property

What You Need to Know

Key takeaway #1

The preliminary Report takes the position that "making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access, goes beyond established fair use boundaries."

Key takeaway #2

Given the recent change at the head of the Copyright Office, there may be further revisions to this guidance in the future.

On Friday, May 9, 2025, the U.S. Copyright Office released the third (pre-publication) installment in a series of reports regarding the intersection of artificial intelligence (AI) and copyright law.1 This report addresses the legal implications of training generative AI models using copyrighted materials.2

Of particular concern to authors is whether they are entitled to license royalties by those who ingest their copyrighted works into AI models for training purposes.3 AI software developers, on the other hand, argue that licensing will unnecessarily restrain development and innovation, and impair competitiveness.4 This third report analyzes those differing views by assessing that the author's right to control unauthorized reproduction is likely implicated,5 and by comprehensively reviewing four non-exclusive factors that should be considered in determining whether use of a particular copyrighted work is fair use:

  1. The purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;
  2. The nature of the copyrighted work;
  3. The amount and substantiality of the portion used in relation to the copyrighted work as a whole; and
  4. The effect of the use upon the potential market for or value of the copyrighted work.

The report concludes that the first and fourth factors may carry more weight in this analysis than the others,6 but that a fair use determination "requires balancing multiple statutory factors in light of the relevant circumstances" on a case-by-case basis.7 For example, while some uses of copyrighted works for AI training may be transformative by "add[ing] something new, with a further different purpose or different character" so as "to advance the purposes of copyright,"8 "making commercial use of vast troves of copyrighted works to produce expressive content that competes with them in existing markets, especially where this is accomplished through illegal access, goes beyond established fair use boundaries."9

The report suggests that a market-based approach would provide a better alternative to government intervention on the issue of whether and to what extent creator's rights prevail over fair use rights and vice-versa. For example, "[p]remature adoption [of statutory, compulsory licenses] ... risks stifling the development of flexible and creative market-based solutions...and can take years to develop."10 Likewise, technological developments in AI may require fewer unlicensed works without sacrificing quality.11

Although the report is preliminary and carries no legal effect on current or future litigation, the comprehensive analysis in the report provides ample support for those on both sides of the issue of whether training AI with copyrighted works constitutes copyright infringement or fair use. As it stands, the report is currently the only guidance on the topic from the Copyright Office. However, the new administration has implemented a change in the leadership of the Copyright Office after issuance of the report, and whether that change will affect this guidance in the future remains to be seen. Given the potential implications for AI developers, content owners, and licensees, we will continue to watch this issue closely as both the legal landscape and technological developments evolve, and we look forward to evaluating how these issues may affect your IP strategies, licensing arrangements, or risk exposure.

Footnotes

1. The first report, released in July of 2024, focused on digital replicas and deepfakes (https://copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-1-Digital-Replicas-Report.pdf). The second report, released in January of 2025, addressed the copyrightability of AI-generated works and the human authorship requirement (https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf).

2. https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf

3. See, id, pp. 32-33.

4. See, id, p. 34.

5. See, id, pp. 26-30.

6. See, id, p. 74.

7. Id, p. 107.

8. Id, p. 37.

9. Id, p. 107.

10. Id, p. 104-105.

11. See, id, p. 105.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More