“Success! Your registration form has been submitted successfully. If you are indeed a member of the class, you will receive formal notice later this year.”
That’s the message you get when you register to receive your personal piece of the $1.5 billion that Anthropic, maker of the popular AI chatbot Claude, has proposed to pay authors after the company was found to have illegally acquired pirated books from notorious online “shadow libraries.” The deal, announced last week , would cover roughly 500,000 pirated titles, with rights holders set to be compensated $3,000 per work — hence that eye-watering $1.5 billion overall price tag.
The terms, if finalized, would make for the largest publicly reported copyright payout in U.S. history and a benchmark for other AI cases pending against OpenAI, Meta, Microsoft, Apple, and others.
But for authors — at least for me — the news feels a bit like buying a scratch ticket, getting excited as the cherries seem to be lining up, and then finding out that you won $60. Maybe. A moment of glee followed by a long, sobering look at the fine print.
If you’ve only scanned the headlines, you might think the ground has been cleared. But in reality, where we’ve landed looks more like the beginning of an administrative nightmare.
At a hearing on Monday, Judge William Alsup called the deal proposal “nowhere close to complete,” postponing preliminary approval and insisting on basic mechanics before any checks go out to the likes of me: a definitive list of covered works, a feasible notice and claims process, plus clear allocation rules for when multiple parties share ownership.
In other words, the court wants a real, workable, detailed process for distributing the funds — and this within an industry, traditional publishing, in which it’s often the case that several parties have claims to rights. Authors retain copyrights to their books, but get publishing deals by licensing those works to companies like Macmillan, Penguin, and Hachette (my publisher). Beyond that, you may also have co-authors, work-for-hire agreements, estates, illustrators, and translators involved.
The bottom line here? The checks are very much not in the mail, and you shouldn’t hold your breath.
Then there are the complex details of the judge’s legal findings and the mixed feelings they’re inspiring across the industry. Alsup’s original ruling, which landed this summer, effectively split the legal baby . He held that training on lawfully obtained books is fair use — a major win for AI developers — while leaving authors’ claims regarding illegally acquired datasets (those from “shadow libraries,” etc.) very much alive. This paints the problem as dirty sourcing, not the unauthorized use of the underlying material.
The proposed settlement is meant to resolve the piracy angle: Anthropic would delete and pay damages for the unlawful copies, while the deal preserves future claims about model outputs if Claude reproduces protected text. But even if this settlement goes forward, other courts could still reach different conclusions about AI training itself, remaining open to the argument that transformative training on lawfully acquired materials is fair use.
So for authors who object to unauthorized use of their books for AI training, the issue is far from resolved, and if anything, looks set back, even as dozens of other cases wind their way through the legal system.
Nor is every author happy with the idea of receiving $3,000 per work as compensation. “Authors tend to say [the proposed settlement] is not punitive enough and no amount of money will ever be sufficient,” Jane Friedman, a widely respected publishing-industry analyst, wrote in her newsletter, The Bottom Line . “Others, especially those who argue AI training is fair use, say authors and publishers have hit the jackpot and that little or no market harm has been done.”
One of the judge’s biggest questions is who exactly is in the class — an interest that I share.
For now, eligibility turns on whether Anthropic (allegedly!) downloaded a given book and whether that book was registered with the U.S. Copyright Office within certain relevant timeframes. From what I can tell, I’m eligible. My book was published and registered for copyright by my publisher in 2021, one of the years covered, and my book has also turned up in similar cases , such as the one against Facebook parent company Meta. But as yet, there seems to be no firm way of knowing. Many authors have no idea .
Worse, some authors are discovering an even more painful wrinkle. One thing the Anthropic case has made clear is that some publishers never registered for copyright of some of their titles — or if they did, did so too late. The plaintiffs have been ordered to deliver a final list of works and a fleshed-out plan in mid-September, with another hearing shortly after . If the judge is satisfied, preliminary approval could still happen this fall. If not, a trial is set for December. If you’re still following me here, and you don’t have a law degree, you’re probably feeling dizzy, the details whizzing, gnat-like, around your brain. Understandably, that’s how a lot of authors and observers feel, too.
Who doesn’t want a random $3,000 check to materialize in their mailbox? I know I do. But here’s why that’s unlikely to happen. The figure is a before-everything number. “Before everything” means it doesn’t include class-action fees, expenses, author-publisher contract splits, potential taxes, and probably some other figures I’m not even aware of.
In other words, the $1.5 billion proposed settlement figure looks substantial. Heck, it may even be substantial. But to get a slice, a lot of pieces need to click together first, and the net number may be a lot closer to your winnings from PTA bingo night.
Believe me, I’m not complaining. As I’ve written before , realizing one’s lifelong author dreams tends to be a long, ongoing lesson in humility — less the stuff of A Star Is Born, more like a combo of Office Space and The Wrestler.
For the publishing industry, however, the overall figure is consequential. Until recently, it looked as if large-scale book piracy by AI companies would go unchecked. Now there looks to be some legal template for protecting rights holders. And that’s not nothing.
Leave a Comment
Your email address will not be published. Required fields are marked *