Authors Should Search Their Works in the Anthropic Class Action Settlement
A $1.5 billion class action settlement will end litigation by authors that alleged AI developer Anthropic used copyrighted written works to train its “Claude” family of large language models without the proper licenses or consent.
The ultimate amount an author or publisher may receive from the Anthropic settlement will depend on how many works they own that are covered by the deal, the total number of valid claims that are filed, and whether multiple class members submit a valid claim form for a particular work.
The settlement website www.AnthropicCopyrightSettlement.com stipulates that if a covered work has more than one legal and/or beneficial copyright holder, the payment for that work will be distributed equally among all copyright holders.
U.S. copyright law states that willful copyright infringement can lead to statutory damages of up to $150,000 per infringed work. Anthropic has also agreed, as part of the settlement, to destroy all files it downloaded from LibGen and PiLiMi, as well as any copies thereof, except scanned copies.
A hearing is scheduled for April 23, 2026 to determine whether the settlement will receive final approval from the court. Payments will begin to be distributed only after final approval has been granted and any appeals have been resolved.
How to Search for Your Works and File a Claim
Search the Works List to see if your books are included.
Submit a Claim Form at www.AnthropicCopyrightSettlement.com to receive a payment of about $3,000 per work before fees if your work appears on the list.
The deadline is March 23, 2026.
You can file now without waiting for notice if you find your work on the Works List.
If you submit a Claim Form online, you don’t need to resubmit the form after you receive a formal notice.
If no claim is submitted by the author, their heirs, or publisher for a title, then you will waive your claims against Anthropic, and will neither receive the funds nor be entitled to sue Anthropic separately.
If you want to be able to sue Anthropic separately, you must opt out.
The Backstory to the Lawsuit
The lawsuit claims that Anthropic infringed the copyright of hundreds of thousands of books by downloading them from pirated databases, Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi), in order to train its generative AI language models.
To be covered by the settlement, any work found in the applicable LibGen or PiLiMi versions must have an International Standard Book Number (ISBN) or Amazon Standard Identification Number (ASIN) that was registered with the U.S. Copyright Office within five years of publication, and was either registered before August 10, 2022 or within three months of a work’s publication.
"This landmark settlement will be the largest publicly reported copyright recovery in history," the settlement motion states, arguing that it will "provide meaningful compensation" to authors or legal copyright holders and "set a precedent of AI companies paying for their use of pirated websites."
Judge Held That AI Training is Fundamentally Transformative as Part of a ‘Fair Use’ Analysis
While some see the settlement as a victory in authors’ fights against AI companies infringing on their intellectual property rights, others want to clarify that the case shouldn’t be seen as a validation of the AI companies’ actions.
Representatives of the Authors’ Guild stated, “Although we respectfully disagree with Judge Alsup’s holding on the fair use question—that AI training is fundamentally transformative—and believe that market harm was not correctly factored into the analysis, we nonetheless see the outcome of this case as a positive step towards protecting authors. We are looking to other ongoing cases, such as our class action suit in the Southern District of New York against OpenAI, to prove that the unlicensed use of books by AI companies to build models that produce competing works is not fair use.”
Right to Train?
There is no legal precedent for AI companies to claim that their LLMs have the right to train off of copyrighted works. Online publication The Trichordist explains, “That analogy collapses under scrutiny. First of all, humans typically bought the book they read or checked it out from a library. Humans don’t make bit-for-bit copies of everything they read, and they don’t reproduce or monetize those copies at global scale. AI training does exactly that — storing expressive works inside model weights, then re-deploying them to generate derivative material.”
Educational Works Not Covered
Default splits are not provided for works published by educational publishers, since there are no standard splits for infringement proceeds in most of those contracts. Splits for those works will instead be determined by any applicable terms in the contracts with the claimants making a good-faith representation of the splits.
For more information on how to complete a claim for educational texts, visit the Textbook and Academic Authors Association.
The contents of this article are intended to convey general information only and not to provide legal advice or opinions, and should not be construed as, and should not be relied upon for, legal advice in any particular circumstance or fact situation. Nothing in this article is an offer to represent you, and is not intended to create an attorney-client relationship.