Last Updated on December 15, 2025, 9:35 am ET
Have you heard of the Library Copyright Alliance (LCA)? We are the voice of the library community on copyright policy. Our founding members—the Association of Research Libraries (ARL) and the American Library Association (ALA)—represent over 300,000 information professionals and thousands of libraries. Together, we work to advocate for balanced copyright policy and fair use so people can continue to access copyrighted works for scholarly and nonscholarly uses.
LCA Position on Copyright and AI
In 2023, when the US Copyright Office launched its study on copyright and AI, LCA developed policy principles stating our fundamental position: that the Copyright Act is robust and flexible enough to address issues at the intersection of copyright and AI, and that no amendment to it was needed. Further, we stated that training AI models on copyrighted works is likely to be fair use, based on our interpretation of the statute and case law, like Google Books, in which a court held that copying works to create a non-expressive database is transformative fair use.
We built on those initial principles in comments to the Copyright Office, and explained that legislation requiring licenses in order to use published materials in training AI models would interfere with private ordering. At the time, we were particularly concerned about legislative proposals that would require licensing or other forms of permission to use copyrighted works in generative AI training. Fortunately, those proposals did not advance. And, in the third part of its report on Copyright and Artificial Intelligence—issued as an unofficial pre-publication—the Copyright Office acknowledged that, while voluntary licensing schemes can continue to develop, legislation mandating licensing is not necessary.
Revisiting Our Statement
Two years later, the intense public interest in copyright AI, the explosion of litigation, and the early court decisions prompted us to revisit our policy principles. This review allowed us to reflect on whether the evolution of the legal landscape is consistent with our interpretation of law. As of today, courts have weighed in on two AI copyright cases—Bartz v. Anthropic and Kadrey v. Meta—in which authors sued AI model developers for training on their content. In October 2025, LCA issued an updated statement reiterating our fundamental position, and reflecting what courts have said in both cases: training AI models on copyrighted works is highly transformative.
The statement also acknowledges the labor disruption that generative AI (GAI) is likely to cause, observing that libraries and others can support AI literacy and retraining workers, while cautioning that copyright law should not be treated as a means for addressing these broader societal challenges.
A number of issues remain unsettled by the courts, and they are not referenced in the statement.
First, both the Kadrey court and the Copyright Office raised the issue of “market dilution,” a theory that could restrict the use of AI for researchers and other authors. Under this theory, outputs of large language models (LLMs) can dilute the market for works that are similar in style to those used in a training dataset. This theory is not based in copyright law, and distorts the interpretation of the fourth factor of a fair use analysis—market harm—which examines whether a work is a direct substitute that people would buy instead of the original work. Although the revised statement did not address market dilution directly, it noted that “at this time, the overall impact of GAI on future creativity is speculative.”
Second, in Bartz v. Anthropic, Judge Alsup suggested that fair use did not excuse Anthropic’s training on pirated works. A trial on the pirated copies used to create Anthropic’s central library and the resulting damages was scheduled for December, but Anthropic settled its lawsuit for $1.5 billion. On the other hand, in Kadrey v. Meta, Judge Chhabria held that Meta’s use of copyrighted works to train Llama was a fair use, even when Meta obtained those works from piracy websites. Legal scholars like Ed Lee, professor at Santa Clara Law, have argued against the idea that pirated books are “irredeemably” infringing, noting that this concept has no basis in copyright law. Currently, there has been no final judgment on whether using unauthorized copies for model training is fair use.
The LCA statement acknowledges what we have suspected about remuneration schemes—that they would likely provide limited revenue to individual creators, given the inherently large volume of works ingested to create large language models. The Bartz v. Anthropic settlement confirmed just how complicated it is to identify rightsholders.
It is worth noting that the cases that have been decided by courts are in the commercial context. Commercial v. noncommercial uses have a significant distinction in copyright law—teaching, scholarship, and research are favored purposes under the fair use doctrine. LCA will watch what courts say about noncommercial uses, and the implications for AI-based scholarship.
Finally, as generative AI raises concerns about labor practices, impacts on the environment, education, and society, we need to refine the way we talk about copyright law without simply waving away these concerns. While copyright law is not the tool to address these challenges, we encourage policymakers and advocates to consider other areas of law that might be more appropriately refined to address issues faced by creators in the context of generative AI.
On Creativity
A challenge with any new technology, and especially with AI, is that it is easier to imagine how existing technologies, their uses, and resulting artifacts might be displaced. But generative AI also has ample positive potential along with its broad applicability. We should expect that—as has been seen with photography, the VCR, the internet and web, and Adobe Photoshop—new forms of creativity will emerge now that easy-to-use AI tools are becoming broadly available. Some new uses might open up new markets that are adjacent to but different from familiar ones, and that development will alter the fair use analysis. And some will enhance creativity by adding new dimensions because doing so is now easier (for example, in an instance familiar to one of the authors, by asking a chatbot to write a poem where only professional prose might have been used before).
Looking Ahead
LCA remains committed to ensuring that copyright law serves its constitutional purpose: promoting the progress of science and useful arts, while protecting the rights of creators and the public interest to access and build on knowledge. As the legal landscape continues to evolve, we will evaluate our statement to ensure that it reflects our interpretation of law. LCA will continue to rely on the statement as the basis for our advocacy on AI and copyright to the US government, and to guide our participation in international coordination and setting of policy relating to AI and copyright.
Marjory Blumenthal is a Senior Policy Fellow with ALA. Katherine Klosek is the director, Information Policy and Federal Relations, ARL.
