Fair Use and AI: What Bartz v. Anthropic Means For Copyright Law & Large Language Models
Author: Beau Reeves
In a groundbreaking decision issued on June 23, 2025, the U.S. District Court for the Northern District of California ruled in Bartz v. Anthropic PBC that using copyrighted works to train large language models (LLMs) qualifies as fair use under U.S. copyright law. The ruling marks a major legal development as courts and lawmakers continue to grapple with how copyright law applies to artificial intelligence.
What Is Fair Use?
Fair use is a legal doctrine that permits limited use of copyrighted material without permission from the copyright holder in certain circumstances. Codified in Section 107 of the Copyright Act, it considers four factors:
Purpose and character of the use – including whether the use is commercial or nonprofit and whether it is “transformative,” meaning it adds new expression or meaning to the original.
Nature of the copyrighted work – factual works are more likely to support fair use than highly creative ones.
Amount and substantiality of the portion used – both the quantity and the importance of the portion used are evaluated.
Effect on the market – whether the use harms the current or potential market for the original work.
Courts weigh these factors holistically, with no single one being determinative.
The Rise of AI and the Legal Challenge
As the U.S. Copyright Office recently noted in its report on AI and copyright, generative AI systems like LLMs depend on enormous datasets—including copyrighted material—to perform tasks such as summarization, translation, and content creation. Whether such training infringes copyright has become a hotly contested issue, with dozens of lawsuits and legislative proposals around the globe.
In Bartz v. Anthropic, a group of authors sued Anthropic, the company behind the Claude AI system, for allegedly using their books to train its LLMs without permission. The authors did not claim the AI reproduced or shared their books but objected to their works being used for training purposes at all.
The Court’s Decision
The court sided with Anthropic on two major points:
Training LLMs is fair use. The court found the use “exceedingly transformative,” emphasizing that the LLMs did not replicate the books but used them to understand language patterns. The plaintiffs could not exclude others from “learning” from their works—a distinction the court found meaningful.
Digitizing purchased books is fair use. Anthropic’s scanning of legally purchased books into a digital library was considered transformative, as it changed the format for internal use without expanding distribution.
However, the court drew a clear line when it came to pirated copies. Anthropic’s use of illegally downloaded works to build its library was not excused by fair use, even if the copies were used for training. Because these pirated works displaced legitimate sales, the court ruled that this use was infringing. This portion of the case will proceed to trial.
Implications
This is the first major judicial opinion addressing fair use in the context of LLM training. It sets a precedent that could influence numerous ongoing cases and signals judicial recognition of the transformative nature of AI training. Still, the ruling reinforces that how data is obtained matters: lawful acquisition supports fair use; piracy does not.
Ultimately, the decision highlights the need to strike a balance—between fostering innovation in AI and preserving a healthy creative economy. As this area of law evolves, courts and lawmakers will continue refining where that line is drawn.
Sources
Anna B. Chauvet & Alyssa S. Mottahed, District Court Finds That Using Copyrighted Works to Train Large Language Models Is Fair Use, Finnegan, June 26, 2025, https://www.finnegan.com/en/insights/ip-updates/district-court-finds-that-using-copyrighted-works-to-train-large-language-models-is-fair-use.html.
Copyright and Artificial Intelligence Part 3: Generative AI Training, United States Copyright Office, May 2025, https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf.
U.S. Copyright Office Fair Use Index, United States Copyright Office, Feb. 2025, https://www.copyright.gov/fair-use/.