It’s fair to say that Margaret Porter of Concord, author of 17 published novels, has mixed feelings about her upcoming payment as part of a massive lawsuit over books that were stolen to help develop more powerful artificial intelligence software.
On the one hand, it’s nice that the A.I. company Anthropic is being made to cough up $1.5 billion to authors and publishers as penance and that she will get a few thousand of those dollars. On the other hand, the settlement avoids the deeper question of whether A.I. companies can “train” their models on created material without paying for it.
Even worse, it does nothing to settle the bigger question of what A.I. is doing to all of the arts by letting people generate vast amounts of material of varying quality with little effort, based on past material created by people.
“It’s nice there will be compensation, but compared to their heinous usage of years of years of work, of research, of struggle to become a published author … it’s not much,” Porter said. “I have seen authors writing on their Substacks or various places where writers congregate. … They’re starting to think, is it worth it? Why am I doing this? Some are really questioning their future and the future of the publishing industry.”
Legitimate use or theft?
The settlement is from a class action suit in which three authors sued Anthropic for downloading books from sites called LibGen and PiLiMi, where people can upload digital versions of books, magazines and other items without needing to get permission or pay royalties. Anthropic used the material to improve its A.I. models.
The lawsuit is part of a broader fight over A.I. companies using art, music, video and other materials to improve the mathematical and probabilistic models in their products, without paying anything to the creators of the material. The companies say this is fair use, allowed under law, while creators say it steals their work and uses it to make products that put them out of business.
“Our creative intellectual property was used to create these engines which are now being used to flood the market of published books,” Porter said. “That’s the most frustrating aspect of the whole thing.”
GEOFF FORESTER / For the Monitor
Porter is one of several New Hampshire authors who are seeking or expecting payments from the settlement.
The 2025 ruling by Judge William Alsup of the U.S. District Court for the Northern District of California did not deal with that issue, which remains unresolved. It punished Anthropic for grabbing books that hadn’t been placed on the two sites legally.
“The settlement was that (Anthropic) downloaded and stored them — piracy, that’s what the judge ruled on. He didn’t rule on whether it’s fair use to take books to enhance the software product — I hate the word ‘training’ — without paying for those books,” said Thomas Tosi, an author and filmmaker whose books are not part of the settlement because they were uploaded later than the court case.
This case is far from alone, as many authors and publishers have filed suits against A.I. companies, including Meta, OpenAI and Microsoft. One website lists 114 active copyright cases against A.I. companies around the country. So far, however, this is the only settlement.
Details, details
Actually filing a claim to get your share of the money isn’t simple, however.
First off, authors need to determine if their works are eligible. Only about 500,000 titles of the 7 million copies of books that Anthropic downloaded from LibGen and PiLiMi meet the legal definition to be part of the class action ruling. There are databases online that authors can use to see if any of their works are among the titles, with the Authors Guild site at authorsguild.org most often used.
Then comes the contentious issue of who owns the rights.
As Porter explained it, the contract between publishers and authors covers the issue of who has rights to the book in the short term and in the long term, and who gets paid if it’s turned into a movie or a manga or an audiobook or is translated for overseas sales. Or, in this case, who gets the roughly $3,000 per title that Anthropic is going to pay.
This is important because many publishers are also claiming some of the money. So, for any given title, the question arises of who owns the rights.
“For my books, and many authors’ books, the rights had reverted to the author for various reasons,” Porter said. “My agents and I were very aggressive in clawing back our rights, because that way we could sell off foreign rights and not have a publisher get a cut of it.
“I was also very meticulous about saving my reversion-of-rights letters … stating that all rights have now reverted to me. A couple of times I have submitted that evidence to the Anthropic-sponsored administrator to make sure they knew that all the money associated with the titles would come to me.”
Writers who didn’t keep those letters or whose agents didn’t keep those letters — or, worse, who didn’t bother to secure their rights in the first place — may see their money pocketed by their publisher or their agent. That is leading to some unpleasant discussions among people who had previously worked together to make art.
“A lot of authors didn’t have their reversion letters. It has been difficult and traumatic for a lot of authors along the way,” Porter said.
There may even be cases where the publisher neglected to file for copyright in the name of the author, which means that, even if a book was printed, pirated to LibGen and then downloaded by Anthropic, it can’t be part of the settlement.
The first portion of the payment is expected mid-November, with the second portion next year.
The bigger picture
The arrival of A.I. that is good enough to quickly create a novel or nonfiction work, even if it’s second-rate or worse, has rattled the entire publishing industry. Amazon, as the most obvious example, has been swamped with A.I.-written books with A.I.-generated artwork to the point that it’s had to limit uploads from each account to two books per week. All that “A.I. slop,” to give its derisive nickname, makes it much harder to find legitimate works by real authors.
The technology has muddied publishing in another way.
A.I. works cannot be copyrighted, which greatly limits how much money can be made from them. So publishers are understandably leery of having anything A.I.-related in books they print.
Writers are having to bend over backward to ensure that they don’t include something generated by a large language model, such as during online research for a scene. Increasingly, they even need to worry about writing that sounds like it might have been generated by A.I., which can draw audience backlash.
“Now publishers are getting very picky,” Porter said. “Authors are nervous that publishers will think, ‘This is A.I.; we can’t publish it.'”
For her part, Porter said, “I have a complete stand-offish relationship with A.I. I’m not using it, not relying on it, other than maybe a spellcheck. It’s just not worth it.”
All these issues are not limited to writers. Visual artists, musicians, video artists — almost any artist with the possible exception of sculptors — are seeing their industry upended at lightning speed. It’s not clear to many what will happen to their audience, to their income, to the whole idea of what it means to make art.
“I am very excited about possibilities coming down the pike for A.I. in scientific fields but I don’t think generative AI has any place in the humanities. By definition, that’s the study of things that are unique to the human mind, human spirit,” Tosi said. “I don’t want to read a book that’s written by an algorithm that never experienced the loss of a loved one, that never fell in love.”
The issue of A.I. and authors will be discussed Saturday, Nov. 14, at a panel session at the 603 Writers Conference at the Grappone Center in Concord.
