Reyansh Sawant and Justin Pae
On October 21, 2026, Anthropic faced a key hearing in a copyright lawsuit brought by major music publishers over the use of copyrighted song lyrics to train its AI chatbot, Claude. The company requested the California federal judge to rule whether using the song lyrics of the other companies in the training of Anthropic’s chatbot, Claude, is considered to be fair use. Universal Music Group, Concord, and ABKCO claim Anthropic’s utilization of the content of theirs, infringes the copyright laws, as well as, the creation of lyrics by Claude harms the market of human-made music. However, there was no ruling given by the judge at the hearing (Brittain, “Anthropic, Music Publishers Square Off”).
What is the obligation of a company toward the makers of creative works, whose works are used in the process of creating a product, which is for the market? The reason why we were attracted to this story is because it connects to our individual interests in the field of technology and in different ways it has impacted our daily lives and routines. As students in the field of Computer Science and Consumer Behavior & Marketplace Studies and in Communication Arts, we were interested in inquiring how AI technology is created and how it could help us get closer to making something creative in our lives. We also seek to study people’s behavior toward innovative technologies in the context of creative making, as well as ways in which the work reaches its audience, via media-circulation. We do not consider music as just mere content; it is deeply personal and a medium for self-expressivity.
The case against Anthropic attracted our attention as it poses some interesting questions regarding the impact of using AI to “generate” something creative.
It goes far beyond the music business. AI technology enables the generation of “quick” ideas, as well as experimenting with various ways of making songs and its lyrics. Nevertheless, when the training on copyrighted material is carried out without any permission of the authors, creators may doubt whose benefit and costs the work produces.
However, Anthropic asserts that the training of Claude through song lyrics is an activity of a completely different nature compared to the dissemination of those songs. According to the company, training AI allows Claude to learn human language and create useful content to be used in various spheres, including science, business, and education. Anthropic believes that the nature of that transformative purpose is the basis of fair use.
There is a commercial conflict here. In their March 2026 the Motion for Partial Summary Judgment argues that Claude has the ability to reproduce copyrighted lyrics and create content that competes with their products or harms the market of these works.
In the October hearing, the lawyer for the publishers indicated that Claude had received more than 119,000 requests for creating “synthetic” lyrics in a six-month period. This number was used as an indicator of market risks. On the other hand, the lawyer for Anthropic noted that there were no specific cases that proved that Claude or any other large language model had decreased the need for original lyrics. The amount of requests does not show how many losses the parties had sustained, and the conflicting positions have not been settled yet (Brittain, “Anthropic, Music Publishers Square Off”).
Another prominent example is the Andy Warhol Foundation for the Visual Arts v. Goldsmith,et al., where the disputants both have reasonable concerns. Utilizing the copyrighted material for training can be beneficial for developing the technology. This technology demands access to existing creative works to generate content through patterns. At the same time, the fact that something is useful does not imply that its development is legal.
The four factors which courts look into while assessing fair use according to Section 107 of the U.S. Copyright Act are as follows: “the purpose and character of the use; the nature of the copyrighted work; the amount and substantiality of the portion used; and the effect upon the potential market for the work. According to the U.S. Copyright Office, the courts take these factors into account collectively and analyze the facts of each individual case rather than applying one general rule.”
Lyrics present an interesting dilemma because they represent a creative endeavor rather than a compilation of information. The AI developer could argue that learning from lyrics allows the model to learn the language and generate new content. The publisher could counter this by saying that the AI system profits from using protected creative content without permission.
There is also the difference between the process of training the model and the outcome of such training. The machine could learn from lots of text without blatantly copying a specific song. However, in the case of generating lyrics of protected work, other issues concerning both the output and its impact on the market appear. It should be noted that the explanation of why the training is helpful cannot solve all copyright issues.
The history of the Anthropic case shows the importance of precision in the legal language. In March 2025, Judge Eumi K. Lee refused to impose a preliminary injunction preventing Anthropic from using lyrics in the future process of training. Such refusal did not mean that the entire process of training used by Anthropic was legal. The October 2026 hearing concerned arguments of the parties at the summary judgment stage (Concord Music Group, Inc. v. Anthropic PBC, Order Denying Motion for Preliminary Injunction).
Asking developers to arrange for individual licenses for each text work used in training might become costly and cumbersome. According to the U.S. Copyright Office, Copyright and Artificial Intelligence, datasets may comprise millions of texts, which belong to diverse people and are licensed differently. For this reason, the problem goes beyond choosing whether to make corporations pay for everything or whether innovation is to prevail over copyrights.
A better solution would be to draw distinctions between training processes, model output, and applications of the generated texts by the AI.
Course Instructor: Dr. Ainehi Edoro
Associate Professor, Dept. of English
email: aedoro@wisc.edu
Editor: Shrinjita Biswas
PhD Candidate, Interdisciplinary Theatre Studies
Works Cited:
Works Cited:
“Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith.” Supreme Court of the United States, 18 May 2023, https://www.supremecourt.gov/opinions/22pdf/21-869_87ad.pdf.
“Authors Guild v. Google, Inc.” Stanford Copyright and Fair Use Center, 16 Oct. 2015, https://fairuse.stanford.edu/case/authors-guild-v-google-inc/.
Brittain, Blake. “Anthropic, Music Publishers Square Off on ‘Fair Use’ Defense in AI Copyright Case.” Reuters, 21 Oct. 2026, https://www.reuters.com/legal/legalindustry/anthropic-music-labels-make-key-arguments-ai-training-case-2026-10-21/.
Disney Enterprises, Inc., et al. v. Midjourney, Inc.. Complaint for Direct Copyright Infringement and Secondary Copyright Infringement. U.S. District Court for the Central District of California, 11 June 2025, https://news.justia.com/wp-content/uploads/2025/06/Disney-Universal-v.-Midjourney-Complaint.pdf.
Brittain, Blake. “US Music Publishers Suing Anthropic Make Their Case against AI ‘Fair Use.’” Reuters, 24 Mar. 2026, https://www.reuters.com/legal/legalindustry/us-music-publishers-suing-anthropic-make-their-case-against-ai-fair-use-2026-03-24/.
Concord Music Group, Inc., et al. Plaintiffs’ Motion for Partial Summary Judgment. U.S. District Court for the Northern District of California, 23 Mar. 2026, https://www.musicbusinessworldwide.com/files/2026/03/Motion-for-Partial-Summary-Judgment-N.D.-Cal.-24-cv-03811-dckt-000594_000-filed-2026-03-23-1.pdf.
Concord Music Group, Inc., et al. v. Anthropic PBC. Order Denying Motion for Preliminary Injunction. U.S. District Court for the Northern District of California, 25 Mar. 2025, https://docs.justia.com/cases/federal/district-courts/california/candce/5%3A2024cv03811/431519/321.
Roberts, Rachel. “‘Training on Lyrics Is Transformative’: AI Firm Anthropic Argues ‘Fair Use’ against UMG in Ongoing Legal Battle.” MusicTech, 23 Apr. 2026, https://musictech.com/news/industry/training-on-lyrics-transformative-anthropic-fair-use-umg-lawsuit/.
United States Copyright Office. Copyright and Artificial Intelligence, Part 3: Generative AI Training. May 2025, https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf.
United States Copyright Office. “More Information on Fair Use.” U.S. Copyright Office, https://www.copyright.gov/fair-use/more-info.html.
Image: Iannis Xenakis. Metastasis (1953–54). Graphic score, © Collection Xenakis Family. Reproduced in Athanasios Argianas, “The Synthetic and Synaesthetic Modernism of Iannis Xenakis,” ArtReview, 24 Oct. 2023.
