Context: Last August, book authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson filed a copyright infringement class action against Claude AI maker Anthropic in the Northern District of California (August 20, 2024 ai fray article). In April, Anthropic filed its opposition to the class certification motion, arguing that a copyright class relating to 5 million books is not manageable and that the questions are too distinct to be resolved in a class action. As part of that argument, Anthropic warned against “[t]he prospect of ruinous statutory damages—$150,000 times 5 million books”: that would mean $750 billion (April 19, 2025 ai fray article). A month later, a hearing was held regarding Anthropic’s motion for summary judgment on fair use, which discussed the extent to which any of the uses of the works in question qualify as “fair uses” under Section 107 of the Copyright Act. During that hearing, Judge William H. Alsup of the United States District Court for the Northern District of California urged the parties to settle (May 23, 2025 ai fray article).
What’s new: The Northern District of California has granted a summary judgment for Anthropic that the training use of the copyrighted books and the print-to-digital format change were both “fair use” (full order below box). However, the court also found that the pirated library copies that Anthropic collected could not be deemed as training copies, and therefore, the use of this material was not “fair”. The court also announced that it will have a trial on the pirated copies and any resulting damages, adding:
“That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for the theft but it may affect the extent of statutory damages.”
Direct impact: This is a mixed ruling on fair use – a loss for both copyright holders and Anthropic, but potentially a big win for AI platforms in general. And, if upheld, the order would mean that AI firms using copyrighted material to train their LLMs may be allowed in the future. The only exception to this would be if the material has been pirated.
Wider ramifications: This is only a summary judgment decision, in the early stages of a case that is currently only being heard in the Northern District of California, with various other AI copyright cases dealing with this same “fair use” issue. It is currently too early to tell how much Judge Alsup will deem appropriate to propose as damages before the jury, although the quantity of copyrighted material Anthropic used in a pirated form was huge, so the damages award could be quite substantial. It should also be noted that the authors could still appeal the unfavorable summary judgment, but this is more likely to happen after the case has gone to trial and a final judgment has been reached.
The class action alleges that Anthropic, which is valued in the tens of billions of dollars and has received backing from Amazon and Google, allegedly copied the authors’ books from pirated and purchased sources, assembling them into its central library to train various Large Language Models (LLMs). This was done without the plaintiffs’ authorization, and Anthropic kept those copies permanently, even if it did not use all of them for training purposes, the authors alleged.
This is the order:
In its summary judgment, the court analysed the question of fair use using a four factors test under Section 107 of the Copyright Act:
- The purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;
- The nature of the copyrighted work;
- The amount and substantiality of the portion used in relation to the copyrighted work as a whole; and
- The effect of the use upon the potential market for or value of the copyrighted work.
Each of these factors presupposes a “use”, so the court had to decide whether the works in question had been used in multiple ways and then evaluate each. In this case, the alleged uses were:
- To train LLMs (argued by both parties)
- To build a “vast, central library of potentially useful content” (argued by the authors)
The authors also argued in their suit that the print-to-digital format change was itself an infringement not abridges as a fair use.
The Northern District of California evaluated each of the uses, finding that the training use was justified as every factor but the nature of the copyrighted work favors this result:
“The technology at issue was among the most transformative many of us will see in our lifetimes.”
While the copies used to convert purchased print library copies into digital library copies were slightly disfavored by the second factor (nature of the work), the court still found “on balance” that it was a fair use because the purchased print copy was destroyed and its digital replacement was not redistributed.
The only issue, which will now be taken to trial, was over the downloaded pirated copies used to build a central library. It ruled:
“Every factor points against fair use. Anthropic employees said copies of works (pirated ones, too) would be retained “forever” for “general purpose” even after Anthropic determined they would never be used for training LLMs. A separate justification was required for each use. None is even offered here except for Anthropic’s pocketbook and convenience.”
Counsel
The book authors’ lead counsel is Susman Godfrey’s Justin Nelson (who also acts for the Authors Guild and at least 28 different authors in a class-action against OpenAI in the Southern District of New York (February 5, 2024 class action complaint)).
Meanwhile, Anthropic’s lead counsel is Arnold & Porter’s Joe Farris.Other cases worth keeping an eye on that are dealing with the “fair use” issue are ThomsonReuters’ copyright infringement suit against Ross Intelligence, which is currently pending in the U.S. Court of Appeals for the Third Circuit (April 16, 2025 ai fray article), and a New York Times copyright infringement suit against Microsoft and Open AI (December 29, 2023 ai fray article).