“`html
AI Training on Copyrighted Books Deemed “Reasonable Use” by US Judge: A Landmark Ruling
Table of Contents
In a potentially precedent-setting decision, a U.S. federal judge in san Francisco has ruled that Anthropic, an AI company valued at $61.5 billion, can train its artificial intelligence models using copyrighted books, whether purchased or not, under the doctrine of “reasonable use” within American copyright legislation. The ruling, delivered on Monday, June 23, 2025, could significantly impact the future of AI development and the ongoing debate surrounding copyright law in the digital age.
The “Reasonable Use” Doctrine and AI Training
The core of the legal battle revolves around the interpretation of “reasonable use,” also known as “fair use,” a legal doctrine that permits limited use of copyrighted material without acquiring permission from the rights holders. This doctrine typically applies to activities such as criticism, commentary, news reporting, teaching, scholarship, and research. The judge’s decision hinges on the argument that training AI models constitutes a transformative use of copyrighted material, justifying its inclusion under “reasonable use.”
Did You Know? The concept of “fair use” dates back to the 1840s and has been refined through numerous court cases, including *Folsom v. Marsh*, which established key factors in determining fair use.
Anthropic’s AI Training practices Under Scrutiny
The lawsuit brought against Anthropic, founded in 2021 by former OpenAI engineers, scrutinized the company’s methods of acquiring and using copyrighted books to train its AI models, including Claude.While the judge acknowledged the potential benefits of AI development, they also addressed concerns about the scale and nature of Anthropic’s data acquisition practices.
The judge stated, “The use of the books in question with the aim of causing Claude has caused spectacular developments and was of reasonable use,” comparing the learning process of an AI to that of a human being reading books. Generative linguistic models, like Claude and its rival ChatGPT, require vast amounts of data to function effectively.
Limitations to the “Reasonable Use” Ruling
Despite ruling in favor of Anthropic on the core issue of AI training, the judge was careful to limit the scope of the decision.The court found that Anthropic’s practice of downloading millions of pirated books to create a permanent digital library did *not* constitute “reasonable use.” This aspect of their data acquisition was deemed a violation of copyright, regardless of the intended purpose.
Pro Tip: Companies should prioritize ethical and legal data acquisition methods,such as licensing agreements and public domain resources,to mitigate copyright risks in AI training.
According to court documents,Anthropic aimed to assemble a library “of all the books in the world” to train its AI models.The judge determined that this ambition, coupled with the use of illegally obtained materials, crossed the line into copyright infringement.
Ongoing Legal Battles and the Future of AI Copyright
The decision is preliminary, and a civil lawsuit will proceed to determine whether Anthropic should be held liable for damages. This case is just one example of the growing number of legal challenges facing AI companies regarding copyright infringement. Musicians, artists, media organizations, and writers are increasingly asserting their rights against AI developers who use their work without permission or compensation.
For example, Sony Music Entertainment, Universal Music Group, and Warner Records filed lawsuits against AI song-generator start-ups Suno and Udio in June 2024 for alleged copyright infringement [2]. Thomson Reuters also sued Ross Intelligence, a legal AI startup, in 2020, alleging copyright violation [1][3].
Key AI Copyright Lawsuits
Company | Plaintiff | Allegation | Status |
---|---|---|---|
Anthropic | Various authors & Publishers | Copyright infringement through AI model training | Ongoing civil lawsuit |
Suno & Udio | Sony Music, universal Music, Warner Records | Copyright infringement through AI-generated songs | Active Lawsuit (June 2024) |
Ross Intelligence | Thomson Reuters | Copyright infringement of legal materials | Partial summary judgment in favor of Thomson Reuters (February 2025) |
the outcomes of these and other cases will shape the legal landscape for AI development and the protection of intellectual property rights. The central question remains: How can innovation in AI be fostered while ensuring fair compensation and recognition for creators?
What are your thoughts on the balance between AI innovation and copyright protection? Should AI companies be required to obtain licenses for copyrighted material used in training?
Evergreen Insights: The Evolving Landscape of AI and Copyright