The Anthropic $1.5B settlement hits home, literally: My author settlement notice just arrived. Once upon a time I wrote some technical books many people bought. Years later, digital copies made their way into the LibGen library used by GenAI companies to train their language models. Like every other author featured in LibGen, neither I nor my publisher gave consent for these works to be used for AI training. Last year, a class action lawsuit against Anthropic, one of the AI companies that used LibGen to train it's models was settled the United States District Court for the Northern District of California. Now every "legal or beneficial copyright owner of a work that Anthropic downloaded from one of two websites: Library Genesis and Pirate Library Mirror" may be eligible to receive a cash payment. If you've ever published something with an ISBN number that got registered with the United States Copyright Office, you may be eligible for this as well. To find out, see the relevant links below. Questions for my author friends and followers: - Did you receive this notice? - Have you registered as a class member? - Do you expect to receive a payment? - Do you believe such a payment will be fair compensation? I'll update as this slow-moving story evolves. -- Class Action website: https://lnkd.in/gyBsN36W LibGen database search provided by The Atlantic: https://lnkd.in/ghWxjucj PBS article on the Anthropic settlement: https://lnkd.in/ggR88ECp Much love to Steve Guttman and his team for building a revolutionary IDE that paved the path to VS Code and the modern age of web development, Anna Ullrich for bringing me in and allowing me to influence its development, and Loretta Yates who took a chance on me all those years ago and guided me through the process of publishing four books! #anthropic #anthropicsettlement #genai #authorrights
This is a pivotal moment, and the 'fine print' of the settlement is just as interesting as the payout. It’s worth noting that while Anthropic agreed to delete the specific pirated source files (like the LibGen/Books3 datasets), they weren't required to retrain the model or 'unlearn' your books. It also sets a fascinating precedent for OpenAI, Google, and Meta. Since they likely trained on the exact same shadow libraries, this settlement establishes a substantial financial benchmark for that specific error. It will be very hard for them to argue 'fair use' on pirated data now that Anthropic has conceded this ground.
Would have preferred to see more codified precedent from the trial route.
What if you were the contributing author to an anthology? How does that work?
Got my notice of inclusion a few months ago. 8 of my books made the hit parade.
As the author of a niche tech book (on pixel shaders and XAML), I honestly have no idea how much of my content made it into AI training datasets. That said, my book is listed among the eligible works, so I qualify for a share of the settlement. I’ve seen the emails regarding the Anthropic class action as well, and my publisher (O’Reilly) has informed me that they plan to file claims for all of their titles. I’m not expecting much from this, especially with the suggested 50/50 split between publishers and authors, not to mention the administrative fees.
I got the notice via postal Mail too, early December. After careful consideration I will Not agree. I find the pro Titel compensation Not Appropriate. It sets a precedent and low standard e g for future licensing negotiations. Likewise I want the leave the option open to sue Anthropic again.
I actually wrote more magazine features and paid web articles (back when that was a viable small income stream) than books. Makes me wonder what kind of legal action is happening on behalf of those publishers.
I have not received a notice. I wrote some books about Joomla that, when I first searched months ago, were a part of the settlement. Then I did a search again and they were not. Perhaps coincidentally, perhaps not: I've started getting residual checks again from at least one of those books. It's not a huge amount by any means. But it's definitely larger than what I would expect for a 20 year old tech book.
Really interesting, looks like my book is in there as well.
As always very interested in Casey Fiesler's take on this, especially the academic publishing angle. For example: What academic publications are considered "works" here, and how would member payout work for a book comprising multiple articles, each with multiple contributing authors?