Anthropic’s $1.5B Settlement: What It Means for Authors (and AI)

2026 will be the year of dumber LLMs.

Anthropic just agreed to pay $1.5 billion to settle a class-action lawsuit [1] brought by authors who accused the company of using illegally obtained copyrighted books to train their models.

But money is one thing. The settlement includes one crucial component that might cause more trouble:

Anthropic must destroy the LibGen and PiLiMi datasets after the expiration of any litigation preservation or court orders.

Now, we’re talking about roughly 500,000 books that came from these pirate sites.

Once they remove all that knowledge, how will Claude perform? And if Meta and others who also used these books follow suit, how will that affect their LLMs?

And this is just the beginning. There are other lawsuits against other AI companies, some of which are or may become class actions [2].

I’ve said it before, and I’ll keep saying it: the AI bubble is real, and it will burst.

I’m not saying LLMs will disappear. No, they’re here to stay. So we’d better get used to the idea of working with AI.
But reality will settle, and things will start to normalize. Being Meta, Anthropic, or OpenAI doesn’t mean you can steal someone’s work.

And by the way, yes, I’m also affected: I found one of my books in the LibGen library. (Should I feel offended that my other books aren’t there? ;))

If you’re an author and want to check whether your book is on the pirate site, you can use this tool from The Atlantic: https://www.theatlantic.com/technology/archive/2025/03/search-libgen-data-set/682094/

And if you believe Anthropic used your book to train their models, here’s the reporting link: https://www.anthropiccopyrightsettlement.com/contact

Note: that doesn’t mean you’ll automatically get $3,000 per book.

Good luck, though!