As artificial intelligence permeates our society, copyright law continues to protect the rights of authors and publishers against infringement by generative artificial intelligence systems and the companies who create them.
Generally, copyright law protects creator’s rights to the works that they create. For published books, authors and publishers determine how a book is produced, distributed, displayed, and if successful, adapted. In return, the author and publisher receives payment for copies of the book or license fees for other uses.
Artificial intelligence companies can use existing books, works, and other documents to train the generative AI models and help the AI system to “learn.” As each company tries to win the AI race, it is incentivized to train their model the fastest to obtain an advantage over their competitors. Content owners allege that not all these companies act in accordance with copyright law, although no definitive court rulings have held that LLMs infringe copyrights. LLMs assert that training their LLMs on copyrighted material is protected by fair use.
Anthropic’s Infringement and Class-Action Settlement
Anthropic, the company behind Claude AI, allegedly copied thousands of books that had been pirated by others to train its large language model (LLM) and promote its multibillion-dollar artificial intelligence business. Instead of purchasing a license to use the books to train its LLM, Anthropic downloaded vast numbers of books from online repositories such as Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi). Anthropic claimed that fair use protected its training of LLMs on the materials. After being sued in a class-action lawsuit, Anthropic agreed to settle the claims against it for $1.5 billion last fall.
It is estimated that Anthropic used approximately half a million copyrighted works without permission. Authors, publishers, and others with rights to those works are now eligible to either submit claims to receive a portion of the settlement payment or opt out and pursue their own lawsuit to seek damages.
Copyright holders who wish to file a claim can do so by March 30, 2026, at anthropiccopyrightsettlement.com. All claims will be reviewed to determine whether the claimant is eligible for payment, and the final payment for each claim is estimated to be around $3,000, although the final amount could vary. While filing a claim does not guarantee payment, it does ensure that the copyright holder receives some relief they are eligible for under the settlement.
If a copyright holder wants to file their own lawsuit to seek damages separate from the class action settlement, they must opt out of the settlement at the same website by February 9, 2026.
The settlement does not prevent authors and publishers from seeking damages on any claims for future infringement by Anthropic or any other AI companies.
What’s Next for Artificial Intelligence and Copyright
The Anthropic case is one of the first looks into how AI and copyright law intersect, but the scope of the case was limited, and the settlement is not binding legal authority.
The case does demonstrate how disapproval of using pirated work to train an artificial intelligence model, and it could be used to set benchmarks for damages caused by pirating works for AI use. But how should AI companies purchase the rights to use books to train the models? Is purchasing a hard copy from a local retailer or an e-book online sufficient to authorize the company to train its AI on that work? The Anthropic case does not provide many answers and further market development or court involvement will be necessary.
The Anthropic settlement also does not address how AI may use lawfully obtained books and other works as part of its responses to queries from AI customers. Can an AI that was trained on a work produce a verbatim copy or near-verbatim copy of that work in response to a request from a user? The level of transformation necessary in such responses to avoid infringement still needs to be determined.
The publishing industry will likely see authors, publishers, and AI companies create comprehensive licensing frameworks that define authorized use while providing compensation to the copyright holders. And surely certain parties will push these frameworks to their limits requiring courts to settle these unanswered questions at the intersection of copyright law and artificial intelligence.
Conclusion
Authors and publishers should consider filing claims in the Anthropic class-action settlement if they believe that their works were used. They should also watch further developments as society and the courts determine how artificial intelligence properly fits within our current markets and laws.
(C) 2026 Bea and Mays. Contact our office for further information at www.bealaw.com