A group of authors involved in a class action lawusuit against Anthropic PBC may soon each get a cash payout of up to $3,000 per work in what lawyers are calling the largest copyright class action settlement in history.  A Federal Court in San Francisco will decide on April 23, 2026 whether to approve the $1.5 billion class settlement.

Anthropic, an American AI research company , founded by former Open AI leaders, is known for its powerful large language model, Claude.ai.

Trial was set to begin in December 2025. The parties then reached a settlement agreement.

Authors and publishers who could be included in the class action settlement have until March 30, 2026 to file a claim for payment. Certain guidelines apply for those who are eligible for inclusion. A “look up” website can help authors and publishers decide if their titles are involved in the infringement. For more information, visit www.AuthropicCopyrightSettlement.com

In June 2025  three authors originally sued Anthropic for copyright infringement after Anthropic downloaded  the authors’ works without permission  from Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi) f0r the purpose of training AI large language models (LLM). The case later expanded to a class action, covering authors and publishers potentially affected.

Anthropic denies all the allocations and any wrongdoing, araguing that its use of the dowbloaded datasets was fair use. The matter was set to go to trial in December 2025, but the parties have now reached a settlement agreement, pending the April judicial approval.

The agreement only covers past conduct (pre-August 25, 2025) regarding specific pirated works; it does not set a permanent, future licensing scheme for all AI training. 
However, authors now have a stronger basis to demand compensation and transparency from AI companies regarding their training data, and a greater incentive to register works with the US Copyright Office. In this case only copyrighted works will be eligible for statutory damages.

The courts distinguished between how data is obtained versus how it is used, finding that training on legitimately purchased books is “transformative” (fair use), but using illegally acquired (pirated) data is not.

The settlement forces AI developers to move away from the “move fast and break things” approach of scraping data, favoring legitimate licensing agreements.

 Companies now have greater incentive to meticulously document their data sources to avoid liability.

The ruling establishes a new framework for AI companies:

    1. Transformation Analysis: Is the use of the data for training transformative?
    2. Provenance Analysis: Was the content lawfully acquired?
Below is added detail from a class action notification to potential claimants.

“The Named Plaintiffs and Anthropic disagree over whether Anthropic has violated the law. The lawsuit has not gone to trial, and the Court has not decided in favor of the Named Plaintiffs or Anthropic. Instead, the Class Representatives and Anthropic have agreed to settle the Action. The Class Representatives believe that the Settlement Agreement offers significant benefits to all Class Members, and that the Settlement is fair, reasonable, adequate, and in the best interest of the Class Representatives and all Class Members.

This class action settlement arises out of a lawsuit brought by various authors against Anthropic PBC (“Anthropic”) alleging that Anthropic violated the Copyright Act by downloading copyrighted works from online websites called Library Genesis (LibGen) and Pirate Library Mirror (PiLiMi) for purposes of training large language models (LLMs). Anthropic denies that it violated the Copyright Act and contends instead that such downloading and subsequent use was fair use under the Copyright Act. The plaintiffs contend that Anthropic’s downloading violated Section 106(1) of the Copyright Act, which gives a copyright owner the exclusive right to reproduce copies, or to license or transfer that right to others. Copying a work without permission is not copyright infringement if a defendant can show the copying was fair use. If the use is determined to be infringement, the Copyright Act provides for statutory damages of between $200 and $150,000 per work, depending on factors including the harm that was actually caused by the infringement, and whether the alleged infringer reasonably believed its use was fair or instead acted willfully. If the use was fair (or there was no copying), the defendant owes $0.

In July 2025, the District Court ruled that the individual plaintiffs (or their associated legal entities) could represent all beneficial and legal copyright owners of the exclusive right to reproduce copies of the works in question. This order was, however, conditional on further steps being taken. Anthropic sought to appeal both the fair use ruling and class certification ruling. The District Court and the Court of Appeals have not ruled on those requests.”

“The litigation involved significant uncertainties. If the case had proceeded, it is possible the class could have been de-certified even before trial, with Class Members receiving nothing. If the case had gone to trial, it is possible the class could have lost on the issue of fair use, again with Class Members receiving nothing. Even if the class could have prevailed at trial and later on likely appeals, a jury might have awarded a wide range of amounts, to be paid only after the added expense and delay of the trial and its likely appeals.

The resulting Settlement is the largest copyright class action settlement in history. It provides approximately $3,000 per work (not per Class Member), plus interest earned on the Settlement fund, less the Court-approved costs and fees taken out of the Settlement fund.

The Class includes legal or beneficial owners of the exclusive right to reproduce copies of a work that was included in the versions of the LibGen or PiLiMi datasets downloaded by Anthropic and is listed in the searchable database called the Works List (described below). A “work,” in this context, refers only to works that:

✔ have an International Standard Book Number (ISBN) or Amazon Standard Identification Number (ASIN);

✔ were registered with the United States Copyright Office within five years of the work’s first publication; and

✔ were registered before being downloaded by Anthropic, or within three months of the work’s first publication.

Class Counsel have compiled a searchable database of books included in the Class and Settlement, available at Works List Lookup, www.AnthropicCopyrightSettlement.com/lookup.

Searching this Works List is the only way to determine whether your work is included in the Settlement.”

#Anthropic, #Claude