Bartz v. Anthropic Settlement: An Update for Authors

The searchable list of eligible works is available at the settlement website (also see the update below): https://secure.anthropiccopyrightsettlement.com/lookup

BACKGROUND

A year ago, a group of authors filed suit against AI company Anthropic over its creation of an enormous library of digitized books to train its Claude LLMs, arguing that because Anthropic did not seek authors’ permission for the creation of the library or for AI training on the included works (claiming, as most AI companies do, that this constituted fair use), Anthropic’s actions constituted copyright infringment. (You can see the original complaint here.)

In addition to purchasing and scanning physical books, Anthropic also downloaded millions of books and other works illegally uploaded to pirate sites. This past June, the judge in the case, Judge William Alsup, determined that while Anthropic’s digitzation of physical books, and its use of these for AI training, did indeed qualify as fair use, its downloading of pirated copies did not.

Accordingly, Judge Alsup certified a class of LibGen & PiLiMi Pirated Books authors (LibGen is pirate site Library Genesis; PiLiMi is pirate site Pirate Library Mirror), consisting of “all beneficial or legal copyright owners” whose work both possesses an ISBN or ASIN, and was registered with the US Copyright Office within 5 years of publication and before being downloaded by Anthropic (August 10, 2022).

On August 26, a preliminary settlement agreement was reached between Anthropic and representatives of the class, with claimants to receive an eye-popping minimum payout of $1.5 billion. The settlement called for the money to be distributed among rightsholders on a pro-rata basis, with payments estimated at $3,000 per work, based on an assumption of approximately 500,000 eligible works.

An official settlement website has been created for authors to provide their contact information to the class counsel (this does not add you to the class; it just makes it possible for you to be contacted with notices about the case).

There’s considerably more detail on all of the above (including how to find out if you may be eligible for inclusion by searching one of the pirate databases) in my previous post about the case. See also the Resources section, below.

IT’S NOT AS SIMPLE AS A CHECK FOR $3,000

Pessimist that I am, I’d expected the per-work payout to be much (much) smaller. But if $3,000 sounds like a lot, and you’ve verified that one or more of your works was included in the LibGen database and are already thinking about what you’ll do with the money…not so fast.

For one thing, as many authors have been discovering over the past weeks, your publisher may not have registered your copyright, even if contractually required to do so. (You can look up copyright registration information here.) If that’s you, you can still register it yourself–but since registration must have been done before Anthropic’s last pirate download on August 10, 2022, it’s too late for you to participate in the settlement.

For another, it could be quite a while before you get anything. Settlements take time to resolve, and the claims process for this one is going to be extremely complicated. The plaintiffs’ proposed timeline (which you can find under the response to Question 10 in this document) sets deadlines for various parts of the process, with a final wrap-up and disbursement date of June 2026 “assuming no appeals”–which is not a safe assumption, nor is it realistic to assume that at least some of the deadlines won’t be extended, especially given the massive task of notifying everyone whose works are part of the settlement. (On the plus side, the settlement funds must be placed in escrow, and will earn interest. So if it takes, say, two years for everything to be resolved, that gigantic pot of money will have grown even larger.)

Additionally…you may not be the only claimant for your work. By defining the class as “all beneficial or legal copyright owners,” the judge included publishers in the settlement (as the Authors Guild explains, “under copyright law, where the author grants the publisher an exclusive license to publish the book in exchange for percentage [of] royalties, the publisher is deemed the legal owner of the publication rights granted to it, and the author is deemed the ‘beneficial’ owner of that right”). Traditionally-published authors with books under exclusive contract will therefore be splitting the proceeds with their publishers. (Publishers showed little interest in the case when it was filed, but as you can imagine, that changed once the settlement amount was announced.)

What about self-published books, or books whose rights have reverted? Especially in the first case, there wouldn’t seem to be any ambiguity about the author being the sole claimant, but I don’t think anything should be assumed at this point. What about books that were reverted and then re-published–might both the former and the present publisher have a claim? Again, nothing is certain–but you can see the complications spiraling outward.

Should publishers even have claims at all? Isn’t AI training a new licensable right that until very recently wasn’t included in traditional publishers’ contracts at all, and therefore was never granted by the majority of authors in the class? Publishers Lunch, which has been keeping a close eye on the case, bursts that bubble (sorry, paywall): the settlement really isn’t about AI training at all.

But the AI training portion of the Anthropic case is over, and sadly the authors and publishers lost. The court ruled that LLM training on this corpus of books was a transformative, fair use.

The charge Anthropic is settling instead of taking to trial is a simple — if gigantic — allegation of plain old copyright infringement. They downloaded millions of illegally pirated works, and according to evidence through deposition and discovery as interpreted by the plaintiffs and, it appears, Judge Alsup, they kept that big old illegal library, for a variety of potential uses beyond LLM training.

THE JUDGE HAS QUESTIONS

To the surprise of many, Judge Alsup quickly rejected the settlement at a hearing on September 8, opining that too many questions on how the settlement would actually work remained unanswered–particularly how the claims process would be handled–and that he was concerned about “hangers on with all this money on the table” (specifically the sudden proliferation of lawyers attached to the case following the class certification).

He also expressed concern that the deal was being “forced down the throat of authors”, possibly by the Authors Guild and the Association of American Publishers lobbying behind the scenes to persuade authors not to object.

Subsequently, he issued two orders, on September 10 and September 16, with a total of 34 questions for plaintiffs and defendant. These focused on a variety of issues, including what happens if there are multiple claims for the same work, what happens when claimants on the same work disagree, what happens if one of several parties entitled to a claim opts out, how claims are resolved if there’s a third party that has licensed derivative rights, how standing to make a claim would be proved, how payments would be divided among claimants, what happens to included works that no one claims…and much more.

Also, per Publishers Lunch,

[Judge Alsup] remains deeply suspicious that the involvement of the Authors Guild, the Association of American Publishers, and multiple law firms suggests “that some kind of global allocation is going to be proposed regardless of the merits of any individual case. For example, the Authors Guild and the AAP might “agree” to a global 50/50 split between authors and publishers. But would it be right and fair to do so with respect to an author whose paperwork clearly shows the author is entitled to all proceeds?”

He warns again, “If a global settlement within a class settlement is what class counsel have in mind, then the district judge will have concern. Please explain the exact and full role of the Authors Guild, the Association of American Publishers, and each of the new law firms.”

On September 22, ahead of a September 25 hearing on preliminary approval of the settlement, plaintiffs and Anthropic filed a supplemental brief that proposed a notification and claims process for class members, including, for trade and university press books, a non-mandatory (i.e., claimants can reject it) 50/50 split of per-work proceeds between author and publisher (this figure is based on typical income splits from rights licensing). While such books represent the majority of works in the class, there’s also a substantial number of other types of work–textbooks, for example–and for those, the splits may be different, depending on contract language.

You can see the sample claim form here.

Also filed by plaintiffs and Anthropic, a day later: responses to Judge Alsup’s questions. Among other answers of interest:

• if a work has multiple claimants, and one or more claimant opts out, the work is excluded from the settlement;
• anyone who does not opt out is bound by the settlement (i.e., their claim against Anthropic is released);
• anyone who opts out can apply for re-inclusion;
• claimants won’t be required to submit documents to prove a claim (on the claim form, they’ll assert copyright ownership under penalty of perjury).

It’s important to stress that these responses are proposed solutions, not necessarily final ones. They may change as the process unfolds. And more questions will surely arise.

A final list of works eligible for the settlement has been submitted to the court, but won’t be released to the public until October. 482,460 works are included (see the response to Question 22).

WHERE THINGS STAND

Yesterday, September 25, Judge Alsup granted preliminary approval to the settlement, calling it “fair” while acknowledging the substantial complications of the claims process.

Representatives of US and international authors’ and publishers’ groups quickly issued statements celebrating the landmark nature of the deal. Remember, though, that Anthropic’s LibGen and PiLiMi downloads are estimated at 7 million works. That fewer than 500,000 qualify for the class should give you a good sense of what a small fraction of this pirated content is actually eligible for redress. And don’t forget that Anthropic is currently valued at $183 billion, and has just completed another round of funding. The settlement isn’t exactly pocket change, but it won’t set the company back significantly.

How are authors feeling about it all? I can only speak for myself, but right now it’s a mix of vindication (some of us won on piracy, at least) and resignation (I fear we may already have lost on fair use).

Still, the fight for copyright must go on. Here’s Association of American Publishers president Maria Pallante:

As it becomes clearer and clearer that one AI company after another has helped itself to the intellectual property of authors and publishers, we hope that courts will recognize that the unlicensed, carte-blanche use of copyrighted works for AI training is not transformative and not fair use. Rather, it flies in the face of the constitutional objectives of copyright law and undermines the full and safe potential of AI for all of us.

“AI companies have an interest and a responsibility to respect and sustain the very industries they rely on to advance their technology. They have a duty to share the tremendous commercial value they have achieved in no small part from the talents and investments of creators.”

As of this writing, there are 51 copyright lawsuits against AI companies, and more will surely be coming.

I’ll be keeping a close eye on the progress of the settlement and providing periodic updates. Watch this space.

RESOURCES

The official settlement website. Currently, you can provide your contact information in order to receive official notices, but the site will be expanded to include information on the claims process, FAQs, and so on–as well as a searchable databse of eligible works.

You may be able to use this resource to see if your works are included in the LibGen pirate database (I’m not aware of any similar resource for PiLiMi).

Look up your copyright registration in the US Copyright Office’s copyright registration database.

SFWA has created a special section of its website devoted to the settlement (applicable to all authors, not just the speculative fiction community). Included is a FAQ, links to resources, and a survey for authors impacted by Anthropic’s illegal downoads. It’ll be updated regularly, so bookmark it and check back often.

UPDATE 10/2/25: The searchable list of eligible works has been posted to the settlement website: https://secure.anthropiccopyrightsettlement.com/lookup.

It’s searchable by title, author name, publisher, and ISBN/ASIN. For me, all my books in the LibGen pirate database are included, except the one my publisher failed to register copyright for. Two of my books didn’t make it into LibGen at all. So of my nine novels, six are eligible for the settlement.

I’m having trouble getting the database to work on my desktop, but it works fine on mobile.

What happens now?

Settlement notices will be sent via email and paper mail to eligible authors for whom the settlement administrator has contact information. Contact info is still being collected, though, and notices will be going out until November 24, 2025. Here’s what the notice will look like.

Notices will provide a unique ID that you can use to file a claim online or by paper mail. If your works are included in the list and for some reason you don’t receive a notice, you can file without the ID. Claims must be filed by March 23, 2026.

The claim form can be seen here. You can submit a single claim form for all your works.

What if you want to opt out, or object to the settlement? There’s a detailed FAQ to address these issues and more, and also a list of key dates by which things need to be done.