Duane Morris Takeaway: This week’s episode features Duane Morris partner Jerry Maatman and senior associate Hayley Ryan with their analysis of the $1.5 billion settlement set to resolve a class action brought by authors alleging artificial intelligence developer Anthropic pirated their work to train its large language model.
Check out today’s episode and subscribe to our show from your preferred podcast platform: Spotify, Amazon Music, Apple Podcasts, Podcast Index, Tune In, Listen Notes, iHeartRadio, Deezer, and YouTube.
Episode Transcript
Jerry Maatman: Thank you, loyal listeners, for being here again for the next episode of our weekly podcast series entitled The Class Action Weekly Wire. I’m Jerry Maatman, a partner at Duane Morris, and joining me today is my colleague, Hayley Ryan. Thanks so much for being here on the podcast today.
Hayley Ryan: Great to be here, Jerry. Thanks for having me.
Jerry: Today, we’re discussing one of the most significant class action settlements ever, a copyright class action settlement that was recently approved for $1.5 billion between Anthropic and a class of authors and copyright holders whose books were allegedly downloaded from online piracy libraries and used in connection with AI training. The court, in its settlement approval order, called it the largest copyright class action settlement in the history of American jurisprudence. Hayley, let’s start with the basics. What was the case about?
Hayley: Yeah, Jerry, so the case was brought by authors and copyright holders against Anthropic. Plaintiffs allege that Anthropic downloaded hundreds of thousands of copyrighted books from the online repositories Library Genesis, or LibGen, and Pirate Library Mirror, often referred to as PiLiMi. The settlement class included copyright owners of books appearing on a specific works list, containing approximately 482,000 works that Anthropic had downloaded from those sources.
Jerry: As I understand it, in the history of the case, the judge, a federal district court judge in the Northern District of California, granted final settlement approval on July 20, 2026.
Hayley: That’s right, Jerry. The court concluded that the settlement satisfied Rule 23’s requirements and was fair, reasonable, and adequate. The court viewed the settlement as providing substantial value to the class. According to the court’s order, the average recovery is expected to be approximately $3,000 per word. The court stated that this amount is roughly four times the statutory minimum damages available for ordinary copyright infringement, and about 15 times the minimum for innocent infringement.
Jerry: As is inevitable in large class action settlements, I understand that some objectors lodged objections to the settlement, claiming that the settlement amount should have been larger, is that right?
Hayley: Yes, they sure did, Jerry. Several objectors pointed out that copyright law allows statutory damages of up to $150,000 per work in certain circumstances. The court concluded that continued litigation carried substantial risks, including the possibility of losing at trial, appellate reversals, delays, and constitutional challenges to a massive statutory damages award.
Jerry: One thing that stood out to me was the participation rate among class action members. I’ve seen settlements where the participation rate is under 3%. What did their data analytics look like in this particular settlement?
Hayley: Yes, Jerry, so here, the participation rate was extraordinarily high. By April of 2026, claims have been submitted for more than 91% of the works on the works list. The court noted that this level of participation was far above what is typically seen in class actions. The settlement administrator received only 350 timely opt-outs, covering just 1,802 works, despite notice reaching hundreds of thousands of potential class numbers. The court also noted that only 54 objectors or comments were filed, and several class members affirmatively supported the settlement.
Jerry: Notice program in a class action settlement is always important, but let’s talk about it in the context of this case, where there were an enormous amount of authors and copyright holders.
Hayley: Yes, Jerry, so the notice program was extensive. The settlement administrator gathered contact information from publishers, authors organizations, copyright records, ISBN databases, commercial sources, and internet searches. Notice was sent to nearly 595,000 potential class members, and the court found that notice reached owners associated with more than 99.5% of the works on the works list. The campaign also included industry publications, newspapers, press coverage, newsletters, as well as social media outreach.
Jerry: The scope of a release in a class action settlement review process is always important. How did it play out in this particular case?
Hayley: Absolutely. So, the settlement only releases claims related to works appearing on the specific works list, and only for certain past conduct. The judge emphasized that class members are not releasing claims concerning future conduct. They’re also not releasing claims based on AI outputs. So, the settlement focused primarily on alleged downloading and copying of works that occurred before August 25, 2025.
Jerry: Class action settlements of this ilk also often involve injunctive, programmatic, or non-monetary relief. How did that play out in the settlement approval process?
Hayley: Yes, Jerry, so Anthropic agreed to destroy the pirated copies of class members’ works that had been downloaded from LibGen and PiLiMi, subject to preservation obligations and other legal requirements. The court viewed that destruction obligation as a significant benefit to the class.
Jerry: And then, inevitably, in all class action settlements, a portion of the amount of the settlement is devoted to attorneys’ fees. I found this opinion on settlement approval particularly interesting insofar as attorneys’ fees approval issues was concerned.
Hayley: Yes, it really is, Jerry. Class counsel initially sought 20% of the settlement fund, which would have been $300 million. They later reduced that request to 12.5%, or $187.5 million. Even after that reduction, the court found the request too high.
Jerry: That’s certainly a significant amount of money to be paid to law firms for prosecuting a class action. Is this because, was the court’s opinion because this is the type of settlement that would fall into the category of what’s called a megafund class action settlement?
Hayley: Exactly, Jerry. The judge concluded that applying a percentage of the fund approach would generate a windfall for class counsel. So, instead, the court used the lodestar method. Class counsel reported approximately $27 million in lodestar fees, and the court applied a multiplier of 3.75. So, that resulted in a fee award of just over $101.5 million, which equates to roughly 6.8% of the settlement fund.
Jerry: Well, an enormous fee, one of the largest of 2026 thus far in any class action settlement, but certainly significantly less than what plaintiffs’ counsel had requested. The named plaintiffs, I assume, also requested service awards?
Hayley: Yes, they did. So, the class representatives asked for $50,000 each. The court found that amount to be excessive, although the representatives devoted substantial time and effort to the litigation, the judge concluded there was no showing of retaliation risk comparable to some employment class actions. So, the court instead awarded $15,000 to each of the three class representatives.
Jerry: Bottom line, Hayley, what are your biggest takeaways for corporate counsel from this decision?
Hayley: So, Jerry, courts remain willing to approve extraordinarily large settlements involving AI-related copyright claims when the settlement provides meaningful and immediate value to the class. However, the fee ruling reinforces that courts will scrutinize attorney’s fees in megafund settlements and may depart significantly from the traditional 25% benchmark.
Jerry: In following, class actions that are filed and settled, I think there’s one more takeaway, maybe, and that’s that this settlement resolves historic claims never before brought involving a defined set of books, but it leaves the door open and invites future litigation involving AI outputs and future conduct in privacy situations. So, it probably won’t be the last major case we see in this particular space.
Well, Hayley, thank you so much for this great discussion and your thought leadership in this area. Thanks for the detailed analysis of this important settlement, and thanks for being here today. Along with our listeners, thank you for joining us for this latest installment of the Class Action Weekly Wire.
Hayley: Thanks, Jerry, and thank you, listeners, it was a pleasure to be here.