Anthropic owes authors $1.5 billion – but the claims process is a mess

Earlier this year, writer Maureen Johnson fought with Anthropic.

Specifically, she struggled with the Anthropic copyright settlement website.

Johnson is the author of 18 books, most of them YA and many of them bestsellers. The AI ​​company Anthropic owes her an estimated $3,000 per book (to be split 50-50 with her publisher) for several of them. The payouts are part of a first of its kind settlement handed down last fall, in which Anthropic admitted to downloading millions of pirated, copyrighted books to train its AI models without the authors’ permission. (According to the New York Times“As part of the settlement, Anthropic said it did not use any pirated works to build publicly released AI technology.”) A judge found that using those books without author permission constituted fair use, but piracy did not. Similar cases are pending against Goal and OpenAI. (Disclosure: Vox’s Future Perfect is partially funded by the BEMC Foundation, whose major funder was also an early investor in Anthropic; they have no editorial input into our content.)

The class action was intended to even the field between individual authors and one of the most valuable companies in the world. To distribute the money to authors, Anthropic and the plaintiff’s lawyers worked with a claims administrator (a company that specializes in managing compensation claims) to set up a website that authors could use to access a small piece of the record $1.5 billion payout.

But Johnson, like other writers who spoke to Vox, quickly hit a problem: The claims website is buggy and unreliable, forcing people to jump through endless hoops to collect the money they’re owed. By March, she had already submitted claims for her 14 eligible titles twice, each time spending 90 minutes painfully filling out the forms.

Now the claims administrator told her they couldn’t find any of her entries. They escalated her through multiple layers of management, each repeating the same thing.

“It became more and more surreal how little this system worked,” Johnson said.

Eventually, Johnson connected with an employee who she said spent the entire call giggling. He told her that he had found her first claim filing from February, but not the new one.

“This system is really messed up,” Johnson said she told him. “It’s just not programmed well.”

Johnson said in response the employee giggled again. “Coding is hard,” he told her.

Johnson is not alone in her frustrating experience. Authors had six months to register their claims for Anthropic’s payout, and many of them struggled to do so.

Anthropic regularly touts its ethical and philanthropic bona fides. (The company is here to serve humanity’s long-term well-being! It’s the safe and responsible AI company! Claude helped NASA’s Perseverance rover travel to Mars!) But the good it does is based on stolen work — and the people who created that work are struggling to get the very small remedy they’re owed.

“Everyone agrees that it’s not the best data.”

All the popular large language models were trained on books; it was the only way to get them enough high-quality text to start generating their own. Most of those books were downloaded from pirated libraries, in at least one case on the grounds that it would simply be too expensive to pay for every title. As it became increasingly clear that this was the case, the class action lawsuits started pouring in.

Bartz et al. f. Anthropic PBC was the first to be established. In September 2025, a judge approved a $1.5 billion settlement between Anthropic and the nearly half a million authors he determined belonged to the class. However, things got tricky when it came time to determine who those half a million writers were.

They had to be authors of books that appeared in one of the three pirated Anthropic databases used in 2021. But trying to create a comprehensive list of those databases proved difficult. Anthropic did not create its own records as it fed pirated books into its training corpus, so lawyers on both sides had to rely on the pirate sites’ own data. And they had to do it quickly, because the trial came with tight deadlines.

“It’s, like, reams of acquired pirated library metadata,” Dave Hansen, executive director of the advocacy group Authors Alliance, told Vox. (Authors Alliance filed amicus briefs in the bartz case and widely published technical explanations for writers.) “I wouldn’t rely on it for almost anything, much less to administer legal claims in a large and important lawsuit. But it was kind of the best they had given the data sources being used.”

“I think everyone agrees that it’s not the best data, but it’s the best they can do on the time frame,” publishing industry reporter Jane Friedman told Vox. “I think that was just the reality for class counsel. The judge really expedited cases, and so they did the best they could in the time they had.”

Neither Anthropic, its attorneys, the class counsel for this case nor the claims administrator responded to a request for comment from Vox. But it appears that the plaintiff’s lawyers and the claims administrator worked together to limit Anthropic’s initial list of 7 million books to only titles that were under US copyright in 2022.

“Then they used a bunch of other industry sources to enrich that data so they had more information about current publishers, and then used that to generate contact information,” Hansen said. “At that scale, it’s really hard to get 100 percent accuracy.” He added: “One of my biggest criticisms of how this settlement and process went is the data. They just weren’t very transparent about it.”

From there, the claims administrator and class counsel used that shaky list to build their flawed website, which is how Maureen Johnson ended up on the phone with a giggling man who told her that coding was hard. Other writers were in a similar boat.

“I have 19 titles in the database,” he said Christopher Moorethe author of funny comedy novels like Lamb: The Gospel According to Biff, Christ’s Childhood Best Friend. After doing the paperwork for 18 of them, he had to step away from his computer. When he returned the next day to finish the paperwork for the 19th book, everything was erased.

A month passed after he submitted the form a second time, Moore said. “And I got another notice: what about these other titles?” Most of the titles belonged to one of the four other Christopher Moores working as authors. One was actually his, Moore said, “but it showed it with some weird Texas copyright.” He filed the claim anyway and is still waiting to hear back.

April Henrywho writes YA mysteries, has also found unusual copyright holders on her books. “One of the books on the list appeared to be an audiobook and showed the narrator as one of the copyright holders,” she said.

Meanwhile, she is struggling to figure out how to handle the seven of her 22 books that she has written with a co-author. “Nobody ever had it in their contract that you were going to split the rights on a legal settlement,” Henry said. “You know what I mean?”

And as writers struggle to navigate the claims process, they do so with mixed emotions.

“That’s not much for your entire catalog.”

Johnson is still furious about his experience with the claims administrator’s website. “Your AI monster ate up all our work,” she said, addressing Anthropic. “Now you’re trying to pay us off with this (…) piece of rubbish that doesn’t work.”

For many writers, the money did not seem enough, as their life’s works were taken without their consent. The full settlement of $1.5 billion sounds like a lot. But split between so many copyright holders, it doesn’t go that far. There’s also the fact that the $3,000 number is just an estimate of what author payouts will end up looking like. In reality, there is a fixed amount of cash available for the class, and the more people participate in the class, the smaller the pot of money available to everyone involved.

“When you think $3,000 a book times 22 books, you’re like, ‘I’m getting $66,000,'” Henry said. But then there is the money that goes to the publishers, and any money that goes to a co-author. “In some cases, it’s going to end up being about $500 per book,” Henry said. “At first you’re like, ‘What a windfall!’ But it doesn’t look like a fluke.”

“To me, that’s an entire career, and it will come down to less than $30,000,” Moore said. “That’s not much for your entire catalog.”

Then there’s the question of what the new world that Anthropic helped build with all those stolen books will look like for writers. “We have no idea what the long-term damage of this is to artists,” Moore said. “I’m on the downslope of my career, so there’s not much they can take from me. But if someone’s strong in the middle of their career, they can really be hurt by this.”

On May 14, the settlement will receive a fairness hearingwhere the judge must review a number of author complaints, including what they describe as “inadequate compensation relative to the harm.”

Meanwhile, Anthropic remains one of technology’s biggest players, currently valued at $900 billion. According to the operating headings: “Anthropic’s Claude claws his way to the top of the AI ​​market.”

Correction, 6 May, 14:00: This story originally misstated the number of books Maureen Johnson had written; it’s 18