I've spent a fair amount of ink on this blog arguing that AI-generated code has no human author, and therefore no copyright. This post is the other side of that coin, and it's personal. Two of the books I wrote were in the pile of pirated works Anthropic used to train Claude, and today I filed a claim in the settlement.
How my books ended up in a lawsuit
The two titles are old friends: Building Web Applications with ADO.NET and XML Web Services and Working with Microsoft Visual Studio 2005 Team System. One is from the early 2000s, the other from the Team System era, and both were properly registered with the U.S. Copyright Office back when I wrote them. I didn't go looking for this fight. The works were swept up in the class because they were among the books Anthropic obtained from pirate libraries to feed its models. When the claims process opened, I checked, found both titles, and filed.
What Anthropic actually did wrong
This is the part worth getting right, because the headline number tends to drown out the legal nuance. In June 2025, Judge William Alsup drew a careful line. Training an AI model on books, by itself, can be fair use when the copies were acquired lawfully. What was not fair use was the piracy: Anthropic downloading and storing millions of books from shadow libraries like LibGen and PiLiMi to build its training corpus. The training wasn't the sin. Getting the material by piracy was. That distinction is what turned a fair-use defense into the largest copyright settlement in U.S. history.
The biggest copyright check in history
Anthropic agreed to pay at least $1.5 billion to resolve the claims, reportedly around $3,000 per title across roughly 500,000 books, with each title paid an equal share after fees. The court gave preliminary approval in September 2025, a fairness hearing is set for 2026, and any distributions will follow after that. I elected a check. I'm not going to retire on two early-2000s programming books, but that isn't the point. The point is that a human wrote those books, a human registered them, and that authorship was worth something a court could put a dollar figure on.
The irony isn't lost on me
Here's the throughline that ties this back to everything else I've written about AI and copyright. My books are exactly what copyright was built to protect: original works of human authorship, fixed and registered. They have real, enforceable value, which is the whole reason there's a billion-dollar settlement to claim against. The code those same models generate on the way out the other side has no human author and, under current law, no copyright at all. Human-authored work went in and was valuable enough to litigate over. Machine-generated work comes out and isn't ownable by anyone. Input and output, two very different stories, and the difference is the human.
If you've ever published a book, check whether you're in the class, because you might be owed a claim you don't know about. And if you take one thing from my odd little brush with this case, take this: human authorship still carries value, in court and out of it. That's worth remembering the next time someone tells you the machine can just write it.