Authors weigh in on $1.5 billion Anthropic AI copyright settlement

This is Gulf Coast Life. I’m Mike Kiniry, thanks for joining us. As generative AI language models have entered the scene in recent years, a wave of copyright lawsuits has arisen in response brought by authors and publishers. These lawsuits hinge on whether downloading and ingesting millions of copyrighted books without explicit permission to train these large language models constitutes copyright infringement, or is protected as fair use. Authors and publishers argue that it is infringement, particularly when AI developers illegally pirate or copy their books to help train their language models. AI companies argue that reading and learning from text is transformative and therefore falls under fair use. In one class action lawsuit that was recently settled, the AI company anthropic, its language model is called. Claude agreed to pay $1.5 billion in a landmark copyright infringement settlement. It's one of the biggest in US history. There are other high profile cases that are similar, including one by publishing houses including Hachette, Macmillan and McGraw-Hill, along with authors including Scott Turow against meta and its CEO, Mark Zuckerberg. That case is still ongoing. The anthropic settlement means payments of roughly $3,100 to the authors and publishers of nearly half a million books, including My Guests Today. Martha Ambrose is an award winning author and English instructor at Florida Southwestern State College. Welcome back to the show, Marty. Hi. Nice to be here. And joining me from Wisconsin is part time Naples resident and full time bestselling author Scott Tarow. Great to talk to you again, Scott. Nice to be with you, Mike. And and with Marty. Yeah. Marty, I'm going to start with you because you're here in studio. Tell us just a bit about yourself and your writing career so we can set the stage for listeners. Okay. I've been writing for about three decades, and I write traditional mystery. Then I moved into historical mystery. I had a book come out 2024, and I have one coming out in September of this year. So I've been a long term writer. I'm sort of a mid list writer, and I supplement that with my teaching career. But it's it's been a great time. So you reached out to me earlier this year to say that you had become a part of this anthropic settlement. I think you said you had four titles. I have, well, this is the catch. I have four titles that meet the criteria. I actually had eight titles that were pirated. Can you explain a little bit about what that criteria is and what the distinction was? Okay. The big distinction was and I know Scott knows this, is that the books have to have been registered with the US Patent Office. Normally your publisher does that. Patent office or copyright office? Copyright office. I'm sorry, the copyright office and normally the publisher does that, but sometimes the publishers don't. And sometimes people who are indie publish. They might not do it for their own books. So for of my books did not get sent in to the US copyright and for did so even though eight were pirated. I'm actually getting paid for four. How do you feel about the way it's all settled out? Well, I have to say I was really delighted because I think it was a big settlement. I think it sends a big message to AI that you cannot pirate books. I'd like to see it move into real change about using it for, quote, fair use training, but hopefully that's going to be step two. OpenAI came out with ChatGPT November of 2022. So coming up on four years ago. Do you remember as an individual when you first had an inkling that maybe stuff you had written were part of it and things like it? Yeah, I didn't notice my work, but it was on some of the writer blogs people were talking about. I think I've seen my work on a book somewhere, and or I'm seeing it used on a website somewhere. And so there was a lot of chatter about it. And I'm a member of the Authors Guild. And that's when I found out, oh, there's a class action lawsuit. So that sort of solidified it. And I've been following it, called all my writer Friends. Every single writer friend I have has had pirated books. Scott, what about you? When we first started getting used to having these language models, when did it first occur to you that the works of yours and authors like you were, were part of the training set? Well, for me, although I've also been a member of the Authors Guild, like Marty and I, actually the president of the organization for different spells and a member of the board. What really clue in was an article in The Atlantic in which really good reporter. Had obtained something from Facebook and explained that most of the LMS, probably all of the LMS employing the silicon theory of it's better to ask forgiveness than permission. He had trained their large language on many of the existing pirate databases of of ebooks that have been that have been obtained without permission of authors or publishers, and that they had then been scraped. Another term of art, in order to educate the LM. You know, when I, I was not clear on the distinction with this anthropic settlement about the the pirated texts versus just copyrighted texts. Isn't pirating books just against the law all on its own, Scott, let alone training a model for it? Well, if you're asking me, the answer is an unequivocal yes. And I, I say this as a former federal prosecutor who every now and then handled copyright cases, and I find the defense that the that the LMS, the biggest companies in the world are offering to be bogus from the start. If if in my days as a prosecutor, somebody had bootlegged a movie and then given it to a distributor and the distributor said, well, yeah, I am using it for a commercial purpose, I'm charging everybody a dollar to see it. But they're all school children, and this is going to be educational for them. My answer would have been, tell it to the judge, you're going to be prosecuted. That's a distribution of copyrighted material that, you know, was obtained in violation of the copyright. And we don't go any further than that under our laws. That's that's actually a criminal violation. So I find the idea that, like the old Yiddish joke, you know, once a chicken, now a duck, that you can somehow say, oh, well, you know, but we really had a good reason to use this pirated material. I don't think the analysis should go any further than saying you're using pirated material. And there's been at least one federal judge who so far has agreed with that. But it's a complicated area of law. And, you know, the one thing we can say for sure about, you know, Google and meta and open AI and and Apple, they've got really good lawyers. So the arguments will be sophisticated and they will go on for many years because who wants to do what anthropic did. And, you know, hand over a chunk of change like a billion and a half dollars. Do you think that billion and a half dollars is like Marty said, it kind of at least it it it's a symbol of pushing back. I mean, do you feel like I want to get into your case in a second? But this anthropic case, do you feel like this is, you know, putting a flag down on behalf of authors and publishers? Absolutely. The great, the great struggle for authors has been to drive home the notion that if you steal work, you've got to pay for it in the end. And that's what the copyright laws require. So I thought that the anthropic settlement was a huge victory for authors and our publishers. There were not everybody agreed with it. There are plenty of people who can explain to you with, with very good logic, that this is a small fraction of what the authors and publishers should actually have received, because books in particular are extremely important for the training of large language models, because they represent the most sophisticated use of language that we have in our culture. And that's what it takes to to make an LLM more sophisticated. So there are plenty of arguments that they should have paid more. And you will hear those arguments if and when we get to settlements with people like Meta and Google. And I don't expect when we get there that other companies will get away for as little as anthropic paid. But bottom line, a terrific triumph for authors to have our rights recognized by a company like anthropic. Marty, it's $3,100, roughly per book. Is that something you split with publisher? Is that something that goes directly to the author? You know, do you know what the details are on that? This is split with the publisher. Yeah. Is it like 5050? Yeah, it's a 5050 because mine is a traditional publisher. The say, indie writer whose published it themselves. They would keep the whole whole amount. When you first reached out to me, you said you had talked to some other authors that you knew. Is this something that, you know a you are all excited by? But B as the case was working through. Is this just something that's taking up a lot of your thoughts right now in terms of this, this world of publishing that's being upended in certain ways? Yeah, we've talked about it a lot. And it's not just a lot of my fiction writer friends. I was talking with a colleague who teaches math at FSU, and as we were talking, I said, you know what? Let me just check something. She wrote two algebra books for Prentice Hall. They were pirated. So it it was so big. 7 million books. So, yeah, it's something we talk about a lot. How are you able to tell that they were pirated? Do you just go to the place where they're pirating them from, or is there some sort of big database that you can verify? Well, the anthropic settlement website has a search for the works, and you can search with the ISBN, or you can search with your name. And I just typed in her name. I said, well, two of them were pirated. There may be people listening to this whose names are on that list. Scott. So you're you're part of the case against meta. And tell us what scribe is. I know that that's somehow involved with this, and I wasn't able to determine what it is. So tell us what that is and we'll talk about the case. You know, for, for various legal reasons, I have a S Corp which owns my copyrights and I called it scribe. So that that's the deep explanation of what scribe is. Is there is that an acronym for something. Oh, I could you know. A lot of letters there. Try to remember. The first words are Scott's copyrights, the S and the C. But and I you know, I think the eyes for intellectual property and books and is the b but, you know, it once was an acronym. I can't even remember what it stands for now. Fair enough. Sorry for asking. So you're part of the case against meta who all is involved? I mentioned a couple publishers. There's other authors. Just tell us the shape of that case, how it differs from anthropic case, if any, and you know how the anthropic settlement may affect it. Well, at the moment, although this can get confusing when you drop down into the weeds. Because venusta. But in this particular case, I am the one author who is in this complaint. And a new complaint that was filed against Google. And basically the reason that I got selected for this task, aside from my role with the Guild, is that, you know, I had another life as a lawyer and litigator, and I, I understand the landscape and how this stuff works. And I am there in those cases at the moment as the lone authors representative. And how does this case differ from the anthropic case? If you ask me, in principle, they're exactly the same. Okay. The the the anthropic case and an earlier case against meta were both pending at the same time. There were 12 plaintiffs in the original meta case. It resulted in one of the more unusual judicial opinions I've ever read, in which Judge Chabrier up in San Francisco, said, well, it's clear what meta did was copyright infringement because, among other things, the LMS can be used to produce competing works with the works that they that they've, you know, munched on. But the plaintiffs didn't argue that, he said. So I've got to dismiss this case. But I note for the record that it's just these 12 plaintiffs who are in front of me, whereas the entire universe of other authors are not bound by my decision. The the case against anthropic was in front of a different judge. And as I said, he didn't get much further in the analysis to say the works were stolen. You use them. That's a copyright violation, which I think, quite frankly, is a pretty good place to start. But as I said, this stuff is going to grind away for a while. This is Gulf Coast life. I'm joined today by novelist Scott Turow and Southwest Florid based author and educator Marty Ambrose. We're discussing the recent resolution of the biggest copyright infringement settlement in U.S. history. The generative AI company anthropic will pay $1.5 billion to authors whose work it used to train its AI models. The suit comes down to the fact that it had used pirated versions of books, not simply that it had used them without permission. It's the first high profile case in this regard that's come to fruition. There are more in the works, including one against meta that Mr. Taro is part of. Are you a part of the other lawsuits? Do you know Marty? I mean, is it is it their class actions? Is there a website where you can search their databases too? I haven't looked at those yet, but I'm hoping. Yeah, I. Don't think at this stage, Mike, until the classes are certified that anybody else is included. It's of course, my fond hope that every single author whose work was scraped illegally will be compensated. And that's from, as far as I'm concerned, the object of both of the lawsuits I'm currently involved in. Is there a simple but still accurate way to describe how that class is formed? In the other cases, it's just the publishers. And you know me as a representative of all authors. Gotcha. I. It's all in the legal weeds to me. We're certainly grateful, Scott. So, you know, as these cases work through Scott, you know what lies ahead. Is there a future where these companies are paying up front to authors openly? Is there a licensing regime that can be developed like like if this works out in the benefit of the authors, what does the future potentially look like? Personally, I would be very much in favor of a licensing regime, which but that could only make apply to works created in the future for the works that have been that have already been pirated. That's a completely different matter. Now, anthropic has said, okay, we're going to pay the authors by a copy of every book that's in this database that's been put together for the settlement, the one where, you know, Marty's got four books and I don't know, I've got 15 or something, and they're going to go out and buy the books. And then they said they'll redo the training. There's certainly no agreement by the authors that they ever that buying a book includes the right to train an LLM on it. And so that'll be, you know, the next frontier. Anthropic may choose to try to seek a license instead of getting engaged in another loot. I, I have no idea. I'm just speculating. And there's, you know, there's a lot of ground left to cover in all of these cases. But you're right to point to the anthropic settlement as a huge landmark in the relations between copyright creators, including authors and the AI companies, because for the most part, they've just sort of, you know, pillaged the available landscape of content, figuring out that they just don't want to wait for fear that they'll fall behind their other AI competitors. This is kind of a can of worms question, and we don't have that much time. But, you know, does copyright extend beyond our US borders, meaning, you know, Chinese language models, other language models out there, you know, if they're doing the same thing, is that just beyond the scope of our law? Well, you're. Asking somebody who is not really a copyright expert. My understanding, of course, is that many of the nations of the world have subscribed to something called the Burn Convention, which under various terms, protects copyrighted works. The Chinese, of course, are not members of that convention, which has created problems in which our intellectual property, that is the intellectual property of American creators, is frequently pirated. You know, and the the Chinese will explain to you that our IP laws are basically capitalist conspiracy to, you know, at the expense of the downtrodden, which allows them to, you know, steal patents and as they are now alleged to have done, take the various LMS and the work of the AI labs, and at least as far as they're claiming, you know, create new AI models that may well spring from the legally protected work done in the United States. Yeah, I've read that they're training their models on the outputs from our models, which were built on copyrighted materials. It's very complicated. Last question for you, Marty. Do you use generative AI for anything in your daily life? I will tell you, I do not. Okay. What about you, Scott? A lot. I've discovered that. It's just an amazing tool. I'm writing a book set in 1963, you know. And if I want to know what gentleman's fashions were in 1963, or whether, you know, certain legal principles, did a person who was not a native English speaker have a constitutional right? Later in 1963, I could have answered that question in the old days, but it would have taken me an hour or two. Now it takes me 20s to get the answer from Claude. Do you feel a twinge of guilt from that? No. I mean, I've actually not. When I'm using Claude, they've paid for. They've paid for what they did. That's true, that's true. Well, I hope you don't mind me, me, me asking that last question. What are you working on now, Marty? You got a book in the works? I do. I have a new book coming out, The Secret Diary of Rome. Which books? That'll be out in September. And, Scott, you got something in your cooking up to? Yeah, I have this book set in for the most part in 1963, which is called Presumed Dead. So should be out next year. All right. Scott Rose, bestselling novelist and part time Naples resident. Thank you so much for taking some time to provide your insights to us, Scott. Mike. Thank you. Always good to talk to you. And Martha Ambrose is an award winning author and English instructor at Florida Southwestern State College. Thank you Marty. Thanks so much, Mike. If you missed any of the show, you can always hear all of our episodes on our website. Our show today was produced by yours truly, with audio production by Matt Krause and video production by David Oquendo. For now, thanks for listening. I’m Mike Kiniry. This is WGCU-FM Fort Myers 90.1 WMKO Marco Island 91.7 FM. We are NPR for Southwest Florida.