Can publishers really represent the interests of authors in AI litigation?
by Kevin Smith
One of the faster moving lawsuits over generative AI has been Bartz v. Anthropic, which is being heard in the Northern District of California by federal District Judge William Alsup. Back in June, Judge Alsup issued a decision on cross-motions for summary judgment. That ruling was a mixed victory, or defeat, for both sides. A really important part of the ruling was that training a large language model (LLM) to use for generative AI was a fair use of copyrighted books. The Judge also found that digitizing legally-obtained books for the purpose of creating an LLM was fair use as well. But he did not find fair use in the use of books taken from so-called “pirate libraries,” holding that the clearly transformative use of LLM training did not immunize the defendant when the titles in its “central library” were collected from an infringing source.
Because of this last part, a clear victory for the authors who brought the lawsuit, there have been settlement talks between Anthropic and those authors, resulting in a proposed settlement which would have Anthropic pay 1.5 billion dollars – a huge amount, but not so much when measured against the over 100 billion capitalization of Anthropic PBC. In any case, the settlement has been stalled by Judge Alsup. The Judge has posed a series of questions for the parties to answer, and it is those questions that have prompted the question in my title.
Judge Alsup has been meticulous in examining the proposed settlement, and his list of questions keeps growing. He started with 16 questions, but in mid-September he added 17 more, for a total, so far, of 33 questions to be answered before he will consider approving the settlement. One interesting aspect of a few of his questions is that he reflects on the possibility that Anthropic might win on fair use for the creation of the central library if that issue went to trial.
To me, one of the fascinating aspects of the Judge’s questions is how they raise the issue of how well the class members in this class-action suit are actually being represented. As is usual in these cases, all of the class members, who are ostensibly represented by the three named plaintiffs, will be bound by the settlement unless they opt out. Judge Alsup asks a lot of questions about how individual authors will navigate the claims process, and how the amount of the payments was determined, in the absence of a court decision. He even challenges whether the plaintiffs will be able to contact the various class members if the settlement is approved. For example, consider questions 21 and 23:
21. If the Settlement Administrator is aware that all individualized notice efforts to a given copyright owner have failed, does the copyright owner remain bound to the settlement?
23. Please explain whatever methods, data, and application of methods to data support the assertions that planned general notice will actually reach 70 to 95 percent of class members or more?
These questions recognize how hard it is to contact a large class, and to bind those authors (meaning they could not sue on their own behalf) if they cannot be reached.
Although the plaintiffs here are individual authors themselves, the Settlement Administrator would, of course, have to work through publishers to deliver settlement monies to the authors represented in this class. Most such authors probably do not know that they are even part of a class-action lawsuit. If they get some money, I am sure most will be pleased. But many will not receive notice or a settlement check because they cannot be found. As both an author and a one-time director of a press, I know personally that publishers struggle to keep up with the whereabouts of many authors. Royalty checks are not sent, or are returned to the publisher, because addresses are out-of-date. Publishers lack the staff or the motivation to take much initiative in keeping up their contact information. Indeed, we hear of so many cases where publishers do not even have copies of contracts with authors, we should not be surprised that they often lack current contact information. So the Judge is right to ask where the confidence of being able to contact 70-95% of the covered authors comes from.
The other question, about whether authors who cannot be contacted remain bound by the settlement, really gets at the issue of whether publishers can really represent the interests of authors. Many authors are probably mad that their works have been used for training an LLM, just as they are often mad about other fair uses. But do all authors want to inhibit this practice? And will the relatively small amount of money each author will get make everyone happy? Some will want more, of course, but they will likely be unable to get it, since they will probably be bound by the proposed settlement. Others may wish the case had not brought for other reasons, including the desire to not allow copyright to inhibit the development of new technologies.
Authors are a highly various group of people, who write for many different reasons. The most wrongheaded thing that Samuel Johnson ever said was that only a blockhead writes for any reason other than money (indeed, it wasn’t true even of him). We know authors write for many reasons, and most get paid little or nothing. But publishers, who are certainly behind this lawsuit, are only in it for the money, and they appear to erroneously believe that all of their authors feel the same. I am glad that Judge Alsup is challenging that notion.
August 28, 2025