HIGHLIGHTS

US Judge Approves Anthropic's $1.5 Billion Copyright Settlement: A Landmark Decision on AI Training and Copyright Law

Introduction

Artificial Intelligence (AI) has transformed the way information is created, analysed, and disseminated. Large Language Models (LLMs), such as Anthropic's Claude, require vast quantities of text to learn patterns of language and generate human-like responses. However, the rapid growth of AI has raised significant legal questions regarding the use of copyrighted materials for training these systems. Authors, publishers, and creative industries have argued that AI developers have relied upon copyrighted works without permission, while technology companies maintain that such use constitutes "fair use" under United States copyright law.

One of the most significant legal battles in this emerging area culminated in July 2026, when the United States District Court for the Northern District of California granted final approval to a $1.5 billion class action settlement in Bartz v. Anthropic PBC.

Background of the Dispute

The dispute originated in August 2024, when non-fiction authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson filed a class action lawsuit against Anthropic, a leading AI research company backed by Amazon and Alphabet (Google's parent company). Anthropic is best known for developing the AI chatbot Claude.

The plaintiffs alleged that Anthropic had downloaded millions of copyrighted books from unauthorised online repositories, commonly referred to as "shadow libraries," without obtaining permission from copyright owners. According to the complaint, these pirated books were used to train Claude, thereby infringing the exclusive rights of authors under U.S. copyright law.

Judge Alsup's Landmark Fair Use Ruling

A defining moment in the litigation occurred in June 2025, when Judge William Alsup issued a significant ruling that drew an important distinction between lawful and unlawful uses of copyrighted material.

The Court held that training AI models using legally purchased books constituted "fair use." Judge Alsup reasoned that the training process was "exceedingly transformative", as the AI did not reproduce or distribute the original works but instead analysed statistical relationships within the texts to generate new and independent outputs. Consequently, the use of legally acquired books for AI training did not amount to copyright infringement.

However, the Court simultaneously ruled against Anthropic regarding its storage of more than seven million pirated books in a central digital repository. Since these books were unlawfully acquired and maintained for purposes extending beyond AI training, the Court concluded that this conduct was not protected by the fair use doctrine and constituted copyright infringement.

Settlement and Final Court Approval

Following extensive negotiations assisted by a third-party mediator, Anthropic and the plaintiffs reached a settlement in 2025, receiving preliminary judicial approval in September 2025.

Had negotiations failed, the matter was scheduled for trial in December 2025, where Anthropic potentially faced damages amounting to hundreds of billions of dollars.

On 20 July 2026, Judge Araceli Martinez-Olguin granted final approval to the settlement, rejecting objections raised by several authors who argued that the compensation was inadequate.

Major Terms of the Settlement

First, Anthropic agreed to pay $1.5 billion, making it the largest known copyright class action settlement in U.S. legal history.

Second, the settlement covers more than 482,000 books, with eligible authors and publishers expected to receive approximately $3,000 per book. This amount significantly exceeds the statutory minimum damages available under U.S. copyright law and is considerably higher than the compensation proposed under the rejected Google Books settlement of 2011.

Third, the participation rate has been remarkably high. More than 91% of eligible authors and publishers submitted claims to receive compensation, demonstrating broad acceptance of the settlement terms.

Importantly, the settlement releases only claims relating to Anthropic's past acquisition and copying of copyrighted works up to 25 August 2025. It does not release claims concerning future conduct or AI-generated outputs. Consequently, authors retain the right to bring fresh legal proceedings if future infringements occur.

The Court also scrutinised the legal fees requested by class counsel. While the lawyers initially sought 20% of the settlement fund and later reduced their request to 12.5%, the Court ultimately approved approximately $101.56 million, representing only 6.8% of the settlement amount. Additionally, 10% of the approved fee was withheld pending submission of post-distribution reports, ensuring accountability in the administration of the settlement.

Objections and Judicial Response

Several authors opposed the settlement, contending that the compensation should have been higher. Others criticised the notice procedures, the limited time available to opt out, and the proposed attorney fees. Certain copyright owners also claimed they had been wrongly excluded from the settlement class.

Judge Martinez-Olguin rejected these objections, emphasising that they failed to reflect the practical risks associated with prolonged litigation. The Court observed that although the plaintiffs possessed compelling legal arguments, a trial would still involve uncertainty regarding liability, damages, and appellate review.  Approximately 350 class members opted out of the settlement, and some have already initiated independent lawsuits against Anthropic that remain pending.

Legal Significance

The Anthropic settlement represents a landmark development in AI and copyright jurisprudence.

Most importantly, it establishes a clear distinction between training AI using lawfully obtained copyrighted works, which may qualify as fair use, and maintaining extensive libraries of pirated copyrighted material, which constitutes infringement. This distinction provides valuable guidance for AI developers and copyright owners alike.

The case is also the first major AI copyright dispute to conclude through a comprehensive settlement, making it an influential precedent for numerous pending lawsuits involving AI companies and creators worldwide.

Furthermore, by limiting the release to past conduct while preserving claims concerning future AI outputs and future copyright infringements, the settlement adopts a narrower and more balanced approach than earlier copyright settlements, particularly the unsuccessful Google Books settlement. This ensures that authors remain protected as AI technology continues to evolve.

Conclusion

The approval of Anthropic's $1.5 billion settlement marks a defining moment in the evolution of copyright law in the age of artificial intelligence. The decision balances two competing interests: encouraging technological innovation through legitimate AI training while protecting the intellectual property rights of authors and publishers.

By recognising that training AI on legally acquired works may constitute fair use, while condemning the use and storage of pirated copyrighted materials, the Court has established an important legal framework for future AI development. The settlement further demonstrates that technology companies cannot ignore copyright obligations and that meaningful compensation can be secured for creators when their rights are violated.

Reference

• Economic Times Legal World
• Reuters
• Associated Press (AP)
• TechCrunch
• Authors Guild
• Reason
• Scottish Legal
• Ars Technica
• NDTV
• Ropes & Gray LLP
• Pinsent Masons
• Business Law Review