What Anthropic’s big payout means for AI and copyright

Summary

---

POLITICO Tech

Episode: "What Anthropic’s big payout means for AI and copyright"

Date: September 11, 2025
Host: Stephen Overle
Guest: Keith Kuperschmidt (CEO, Copyright Alliance)

---

Overview

This episode explores the implications of Anthropic’s proposed $1.5 billion settlement with authors’ groups over the use of copyrighted books for AI training. Stephen Overle and Keith Kuperschmidt delve into whether this signals a new era of accountability for AI companies, the complexities of fair use in machine learning, and the battle lines forming in courts across the U.S. over copyright and artificial intelligence. The conversation also widens to possible outcomes for ongoing and future cases, and questions whether legal action or negotiated licensing is the best path forward.

---

Key Discussion Points and Insights

1. Anthropic’s Settlement: What Does It Mean?

• Summary: Anthropic, an AI company, agreed to pay $1.5 billion to authors whose books were used to train its models—the first settlement of its kind at this scale, though not finalized due to ongoing court scrutiny.

• Kuperschmidt’s stance: The settlement is seen as a positive indication that tech companies can compensate creators without harming their business.

  • Quote [02:05]:
    "It sends a message to AI companies that they're going to be held accountable...they need to be responsible, they need to act ethically as well, otherwise they're going to be subject to serious consequences."

• Financial Impact: Despite the hefty settlement, Anthropic quickly raised much more capital, showing that such payouts don’t threaten their viability.

  • Quote [02:59]:
    "They agreed to the $1.5 billion settlement. They got like an infusion of $10 billion... It's a fly in the back of their neck or something, quite honestly."

---

2. Where Are We in the AI Copyright Legal Saga?

• This is Not the End: The Anthropic case is just the beginning, with around 50 similar lawsuits pending.
  • Quote [03:34]:
    "We're in the top of the first inning, basically, so long way to go."
• Multiple Cases, Diverse Scenarios: Each major AI copyright case involves unique facts and will be decided independently.
• Ongoing Uncertainty: There’s no single precedent set yet; rulings may vary.

---

3. Breaking Down the Legal Issues: “Fair Use” and Illicit Sources

• Key Distinctions in Usage: The case considered three categories:
  1. Training on works legally accessed;
  2. Works Anthropic bought and copied;
  3. Works obtained from “shadow libraries” (illicit sources).
• Court’s Findings:
  • The first two kinds (if legally obtained) were found “transformative” and were ruled fair use by the district court.
  • Use of pirated books from shadow libraries was not fair use and led to liability.
    • Quote [06:17]:
      "With these pirated works that they got right from these shadow libraries... that worked against them... and that's what the judge said. And it could have... lead to billions and billions... in terms of settlement..."

---

4. Is the Settlement a Win or Loss for Creators?

• Some argue it’s a partial loss for authors, since many uses were found permissible.

• Kuperschmidt emphasizes the nuanced nature of the ruling:

  • Illegal use is clear-cut liability;
  • Some uses—even if financially significant—are still fair use under current law.
  • Quote [07:36]:
    "One thing you have to look at... why was Anthropic liable here? Well, they were liable because they went and they copied pirated works from an illicit source. I wish I could say that Anthropic was the only AI company doing that, but they're not."

• Appeals are likely, potentially going up to the Ninth Circuit—or even the Supreme Court.

---

5. Precedent and the Road Ahead

• Limited Immediate Precedent: Each case depends heavily on its facts—no single case will definitively resolve all questions.
  • Quote [10:26]:
    "I don't think any of the cases themselves just standing alone is going to set some type of precedent... All are going to come out a little bit differently depending on the facts."
• Message Sent: There is now a clear risk for AI companies that use pirated content—they will be held accountable.
  • Quote [11:33]:
    "The settlement, I think, sends a message to all AI companies... that they will be held accountable. Because that's exactly what happened here..."

---

6. Is This Opening the Litigation Floodgates?

• Litigation Already Surging: Regardless of Anthropic’s settlement, lawsuits are proliferating—two more filed in the last week alone.
  • Quote [12:10]:
    "Regardless of this case, there have been so many AI cases, AI training cases filed."
• Teaching Moments: New rulings give both creators and AI companies lessons for future legal battles.

---

7. Should the Courts Decide, or Is It Up to Congress?

• Courts as the First Stop: Kuperschmidt believes early court rulings will shape initial expectations.
  • Baseball analogy: we’re early in the “game,” and outcomes may shift as “innings” pass.
    • Quote [13:51]:
      "Yeah, I think the courts are the best venue, at least at this point."
• The Case for Licensing Over Litigation:
  • The guest hopes that settlements will lead to more voluntary licensing deals, not just lawsuits.
    • Quote [16:16]:
      "We're just saying just respect our works, be responsible, be ethical, and let's sit down and figure out a licensing mechanism that works. Licensing is not impossible. It is happening now..."
  • Right now, most major players are committing to litigation, while smaller companies are more likely to negotiate licenses.

---

Notable Quotes & Memorable Moments

• Fairness and Feasibility:
  • [02:05] Keith Kuperschmidt: "It sends a message to AI companies that they're going to be held accountable...and that they need to be responsible, they need to act ethically as well."
• The Baseball Analogy:
  • [03:34] Kuperschmidt: "We're in the top of the first inning, basically, so long way to go."
• On Shadow Libraries:
  • [06:17] Kuperschmidt: "They got them from illicit sources. So that's not good. Right. That worked against them, but they also used them...just kept them in a library. And so that stepped over the line."
• On Licensing as the Way Forward:
  • [16:16] Kuperschmidt: "We're just saying, just respect our works...and let's sit down and figure out a licensing mechanism that works."
• Final Hope:
  • [16:16] Kuperschmidt: "...that is absolutely my hope. It's been my hope for a long time that just come to the table. What they're trying to get is a freebie."

---

Timestamps for Key Segments

• [00:30–01:53] Introduction and context: Anthropic’s $1.5B settlement
• [01:54–02:59] The Copyright Alliance’s view on the settlement’s fairness and impact
• [03:23–04:25] The bigger picture: Anthropic as just one of many pending cases
• [04:25–06:13] Explanation of the legal issues, fair use distinctions, and “shadow libraries”
• [06:56–09:35] Wins and losses for creators; implications of the court’s opinion; possible appeals
• [10:02–11:52] Is the Anthropic case a precedent or a one-off event?
• [11:52–13:24] The proliferation of lawsuits and what these cases teach plaintiffs and companies
• [13:24–14:32] Courts vs. Congress: Where should complex copyright/AI disputes be solved?
• [14:34–16:16] Licensing vs. litigation: Possible resolutions and emerging industry trends

---

Conclusion

The Anthropic case, while splashy with dollar figures, is only the start of an ongoing struggle over AI and copyright. Key takeaways:

• The legal landscape is far from settled—dozens of cases remain to be decided, each with its own circumstances.
• Courts are likely to shape initial outcomes, but negotiated licensing could provide a more sustainable avenue.
• The settlement sends a message, but is not a definitive roadmap.
• Both sides—creators and AI companies—are learning as they go, and the industry remains in its “early innings.”

For now, the fight is on—and its resolution may reshape the boundaries of creativity, technology, and ownership for years to come.

Transcript

[Panasonic Narrator]

For over a century, Panasonic has harnessed the power of innovation to advance industry and improve lives. Today we are continuing that legacy by investing in US Based manufacturing, reinforcing resilient and secure supply chains and creating high quality future ready jobs. Learn more about how Panasonic's efforts are directly supporting economic competitiveness and long term energy security.

[Keith Kuperschmidt]

Foreign.

[Stephen Overle]

Hey, welcome back to Politico Tech. I'm Stephen Overle and on this show I break down tech politics and policy with the people shaping our digital future. You probably saw the recent headline that Anthropic agreed to pay $1.5 billion to authors whose books the company used to train its AI models. It's an eye popping amount and one of the first big settlements of its kind. It's also not a done deal. Earlier this week, a district judge raised doubts about the payout. But the question on my mind is what this settlement would mean for other legal battles. There are a lot of cases hanging out in the courts right now over the use of copyrighted material to train AI. So could this be the first of many big payouts to come? I put that question to Keith Kuperschmidt. He's the CEO of the Copyright alliance, an organization that represents publishers, movie studios, authors and other creators. On the show today, Keith and I talk about the anthropic case and what it tells us about where the copyright fight is going. Here's our conversation. Keith, welcome to Politico Tech.

[Keith Kuperschmidt]

Thank you very much for having me.

[Stephen Overle]

So we're talking about this settlement proposal involving Anthropic. I'm curious, have you all taken a position on this settlement? Are you in favor of what's been proposed here?

[Keith Kuperschmidt]

I don't know that we've taken necessarily a position like we look at the settlement, we look at the terms that we know about and read about and certainly think it's a, it's a fair settlement. We don't know all the inner workings of the case. We think it sends a message to AI companies that they're going to be held accountable. It sends a message that they need to be responsible, they need to act ethically as well, otherwise they're going to be subject to serious consequences. And I think what it does is send a message too that looks AI companies, you can pay, you can compensate copyright owners and you're not going to be out of business, you're not going to go. I mean, the Anthropic is, I think the day or the day after something. They agreed to the $1.5 billion settlement. They got like an infusion of $10 billion.

[Stephen Overle]

So this is a lot of money, right?

[Keith Kuperschmidt]

Fly in the back of their neck or something, quite honestly. And so we don't take a position other than say that, like, I think this is a. This is. This is good news. This shows a lot of the things that we've been saying all along that these AI companies can afford to pay copyright owners. They should be paying copyright owners. And this is a good. This case is a good example of that.

[Stephen Overle]

Let's say the settlement does go through. I mean, are the stakes here that this is now kind of case closed in this battle between AI and copyright? Is that the significance here or not quite so much.

[Keith Kuperschmidt]

I think I and many others wish that were the case, but in reality, this is just one case amongst what is now 50 other cases that are pending or that have been filed, I should say, across the country. So this one is probably furthest along because there's been a district court decision and now a settlement or what appears to be a settlement in the case. But there are many, many other cases across the country that are going to have to be decided, and they all have different facts and different situations. And so what I like to tell people, being a big baseball fan, I'd like to say we're in the top of the first inning, basically, so long.

[Stephen Overle]

Way to go wrong way to go.

[Keith Kuperschmidt]

And one team may feel good about, oh, we got some runs, or we, oh, my gosh, there was some runs scored against us. But there is a long, long way to go here.

[Stephen Overle]

Well, let's talk about both the distance we have to go, but also some of the particulars of this case. And if we take a step back, my understanding here is there's been kind of two big issues involved in this case, one of which was this question of whether Anthropic could use books to train its AI algorithm and whether that was considered fair use, protected under the law. That's something that back in June, the judge ruled was the case. But I understand that's not exactly settled law. There is still dispute around that. What does that mean for creators? Exactly.

[Keith Kuperschmidt]

So what the judge in this case, Bart's Versanthropic, said was he rightfully separated the analysis into the three separate uses that were taking place. One use had to do with the use for training and training only of works that they had legal access to. Okay. Then there was a second analysis having to do with the training that they did on books that, that they bought and then they copied, right? They broke the binding and copied the books, I won't get into too much detail there. And the third issue and the one that has come up in the settlement, the third type of use is the use of these pirated works that were copied from illicit databases they like to call shadow libraries. That's a great term because they're not really libraries, but so called shadow libraries. And so the three different uses, and this last use, which is an issue in the settlement of the case, was held not to be fair use, not to be transformative, not to be fair use. The first use and the second use work were considered by the court to be transformative uses and therefore fair uses. But once again, that's just by a district court that there's no doubt that that would have been and may still be appealed.

[Stephen Overle]

Got it. And so the issue kind of comes down to how Anthropic got these books.

[Keith Kuperschmidt]

It's how they got them and how they used them. So the fact is, with these pirated works that they got right from these shadow libraries, they got them from illicit sources. So that's not good. Right. That worked against them, but they also used them. They just kept them in a library. They didn't use them necessarily for training. They maybe used them a little bit for training, but they also just kept them for a library. And, and so that stepped over the line, if you will. And that's what the judge said. And it could have, you know, lead to billions and billions as it did in terms of settlement, but dollars worth of damages.

[Stephen Overle]

There's some legal commentary I've been reading around this particular case. I'm curious to get your thoughts on. One of which is this idea that, you know, because it was determined that Anthropic used some of these books legally, that this is ultimately a loss for authors and for creators. Even if, you know, this settlement leads to a payout, you know that, that this is still a loss for kind of the creators in this tug of war with AI. Is that how you see it?

[Keith Kuperschmidt]

So I wish it were so simple. Okay, so I mean, just like maybe it's just me being a lawyer, right. And looking at all the crevices and the nuances in all these cases.

[Stephen Overle]

Of course, yeah.

[Keith Kuperschmidt]

One thing you have to look at is the fact that. Okay, and I'll get back to your precise question in a second, but why was Anthropic liable here? Well, they were liable because they went and they copied pirated works from an illicit source. I wish I could say that Anthropic was the only AI company doing that, but they're not. There are many, many other AI companies that are using those same or very similar shadow libraries or illicit sources and using pirated works because that was the cheap way out and the easy way out, and that's what they did. So the fact that Anthropic was held liable, and we don't know by how much, but then ended up settling the case, that is likely to apply in many, many other cases. So that's the first part of your answer. The second part of your answer is the flip side of that though, is that the court in the case did say that, that certain uses by Anthropic were transformative and therefore were a fair use. And that's something certainly that the creative community is not pleased about. Certainly that part, as well as the first part that I mentioned all could go up at some point on appeal to the ninth Circuit, depending on what happens with the settlement. And there's a, I'll call it a sister case, if you will, involving Meta, that's a cadre versus Meta that that was decided two days later, that also in the same circuit will presumably go up on appeal. So this is far from the last word that we will hear on these cases. The ninth Circuit certainly will play a huge role and who knows, maybe even the Supreme Court will play a huge role. And then we have cases in other district courts in other circuits and they'll play a role. So this is, like I said, far, far from clear and far from done at this point.

[Panasonic Narrator]

For over a century, Panasonic has harnessed the power of innovation to advance industry and improve lives. Today we are continuing that legacy by investing in US Based manufacturing, reinforcing resilient and secure supply chains, and creating high quality future ready jobs. Learn more about how Panasonic's efforts are directly supporting economic competitiveness and long term energy security.

[Stephen Overle]

This case has obviously made a lot of headlines in part because it's one of the first big settlements of this nature. Is it actually precedent setting though? Like, does this actually provide any sort of roadmap or any sort of indication of where we're headed in these other lawsuits or on this issue? Or is it really just kind of a standalone case that doesn't offer those things?

[Keith Kuperschmidt]

I don't think any of the cases themselves just standing alone is going to set some type of precedent and therefore decide, okay, wait a minute, this is all fair use or this is all not fair use, or for that matter, oh, okay, since anthropic paid 1.5 billion, that's either the going rate or that's what has to be paid. Because fair use and copyright law is all based on specific facts decided on a case by case basis. And so the facts in, for instance, the Medicase were a teeny bit different. The facts in there's a New York Times case going on. There's a case where with Concord Publishing, there's a whole bunch of other cases which involve very different facts, different scenarios, and they're all going to come out a little bit differently depending on the facts potentially. So I'm not sure how much we can take away from this in terms of its overall impact. One thing I think we can clearly say is that there is going to be this sort of a level of accountability. The settlement, I think, sends a message to all AI companies, or for that matter, anyone else attempting to copy and use pirated works from illicit sources for AI training or other uses, that they will be held accountable. Because that's exactly what happened here, is that the judge looked at what was going on and said, look, this is wrong, this, this, they need to be liable for this, they need to be culpable for this. And that's what happened and that's how we came to a settlement in the case.

[Stephen Overle]

Do you expect more lawsuits to come about as a result of this? I mean, I've seen some commentary sort of saying this is akin to sort of a shakedown of anthropic and that, you know, now we're going to see a lot of folks coming after AI companies trying to get settlements. I mean, is that a reality you expect to see?

[Keith Kuperschmidt]

So regardless of this case, there have been so many AI cases, AI training cases filed. We've had, I think in the last week, 2 of additional cases filed. I don't think this anthropic case spurred those cases on necessarily. I do think the anthropic case, just like the cadre versus medic case, is a teaching moment, if you will. I think AI companies and potential copyright owner plaintiffs are all learning about, okay, how did the court handle this? Like the cadavers, Medicase gives an absolute roadmap for how copyright owners can prevail in their cases against AI companies. The anthropic case does as well. It basically says, like, look, if they're using these pirated, these pirated works, well then you've got something to hang your head on. Whether another court adopts that same reasoning and analysis, I mean, that remains to be seen. We're, like I said, at the very beginning stages, it doesn't seem like there's going to be any decisions, though, in other cases until maybe the spring of next year, if I do have a crystal ball and look into it. So these are the only cases we have for a while.

[Stephen Overle]

As you're talking here and thinking about the idea of all these lawsuits with all these different facts, it raises the question for me. Are the courts even the right venue for ultimately settling these questions, or is it really going to come down to Congress, for instance, kind of acting and updating copyright law for the AI era? I guess I wonder if we're just going to have a million lawsuits with a million different sets of facts because AI is not like it's going away anytime soon.

[Keith Kuperschmidt]

Yeah, I think the courts are the best venue, at least at this point. And the reason I say that is because you're going to have the courts look at these cases on different facts, come out with decisions, and after we get a handful of decisions from different circuits, I think. I think we should have a better idea of where things are going right now. We don't. But like I said, just like the baseball analogy I mentioned, after we get to, let's say, the fifth inning or so, okay, we might get a guess. Okay, the other team's got a lot of runs on the board. Maybe, you know, it's we're going to lose the game or we're going to win the game.

[Stephen Overle]

Right.

[Keith Kuperschmidt]

Maybe it's a tie score at the end of fifth, five innings. We don't know. Maybe I'm bringing that baseball analogy way too far, so I'll stop with that.

[Stephen Overle]

But it makes sense to me.

[Keith Kuperschmidt]

You get some sense that after a bunch of these cases are decided, we all, I mean, no matter what side you're on, no matter who you're rooting for or what your views are, we should get a better sense of where this is going. And it's at that point or even later than that point that, that maybe Congress goes, okay, this is sort of a mess here. We have to figure things out. Or maybe they don't. Maybe it's clear and they don't need to jump into the fray. Our view has always been the best way to handle this, frankly, is not the courts and is not Congress and is not the executive branch is just to sit down with the parties and reach agreement, like these licensing agreements that, I mean, the AI companies and the plaintiffs in these cases are paying these law firms tons of money. Take that money, put it through licensing, come up with licensing agreements that work. And we are seeing that happen. We are seeing that happen in various different contexts. And the interesting thing is it's happening with the little guys, like a lot of the smaller companies. There's certainly some contracts, some licenses with Meta and OpenAI, but a lot of them are with the smaller AI companies as well. And if they can license, certainly the larger, you know, AI companies like a Meta, OpenAI or Google or whatever, they certainly can, frankly. They have the money to litigate these cases, and that's why they're doing it.

[Stephen Overle]

I think I was going to ask you if, you know, one of your hopes coming out of this settlement is maybe it nudges more of those companies to kind of come to the table and say, look, it's just going to be a path of path of least resistance if we negotiate licensing deals and sort of avoid a litany of lawsuits down the road.

[Keith Kuperschmidt]

Yeah, I mean, that is absolutely my hope. It's been my hope for a long time that just come to the table. What they're trying to get is a freebie. We're just saying just respect our works, be responsible, be ethical, and let's sit down and figuring out a licensing mechanism that works. Licensing is not impossible. It is happening now. Let's, let's, let's work something out. But basically they are trying to use sort of be bullies on the block, if you will, and use all the money that they have to just try to prevail in the litigation and maybe that will work for them. But we certainly hope that's not the case. We would like to see a more reasonable approach to these issues, one that doesn't involve so much litigation. But right now, that's the avenue that most copyright owners feel that they're forced to take at point this for.

[Stephen Overle]

Well, listen, Keith, appreciate you being here on Politico Tech.

[Keith Kuperschmidt]

Sure. Happy to be here. Thank you very much for inviting me.

[Stephen Overle]

That's all for this week's Politico Tech. If you like Politico Tech, go ahead and subscribe and recommend it to a friend or colleague. And for more tech news, subscribe to our newsletters, Digital Future Daily and Morning Tech. Our producer is Normal Malaiko Pran Bandy made our music. I'm Stephen Overle. See you back here next week.