My books were pirated and used to train AI

If you use generative AI to put words on paper, I want you to know it’s recently been reported in the Atlantic that at least one tech company – Meta – developed their AI using pirated books and pirated articles. It’s likely other tech giants have done the same.

More, I want you to know that at least one company – Meta again – used pirated copies of my books to train their AI.

How do I know? Well, you may have heard that George R.R. Martin, Jodi Picoult, and a bunch of other authors in the USA have taken a class action against Meta. They’re arguing that Meta broke the law by using their books, without permission or payment, to train its generative AI product. Meta are saying it was fair use to do this. The authors are saying it’s not fair at all. Judges are still deciding.

As part of that class action, Meta had to disclose what training dataset (i.e. what books and articles) it used and where it got it from. Turns out they got the books and articles from a pirate site.

The Atlantic recently published a tool authors could use to see if Meta included their books in its dataset.

Three of my books are there: Mended with Gold, Salt Magic Skin Magic, and Seducing the Sorcerer.

So, setting aside the fact that I didn’t give Meta permission to use my books to train their AI, they didn’t even pay me the few dollars it would have cost them to buy my books. They got them from a pirate site.

If you think I sound angry, you’re right.

The piracy is bad enough, but for a lot of authors, including me, the issues run deeper.

Because, for me, it’s not just the text of those three books that Meta took from a pirate site to train their AI. It’s the decades of work I put into learning how to write, how to have workable ideas, how to edit, how to proofread. It’s the hundreds of hours I poured into writing those books. I sweated over every word. And that’s what I believe they’ve taken – my skill and knowledge, along with the skill and knowledge of thousands of other writers.

Even if Meta are found guilty in this class action, do you suppose I’ll see a single cent?

Let’s be optimistic for a second. Let’s calculate what they might pay me:

1. For the text of the three books they took from a pirate site, they owe me about $12.00 – because that’s how much my books cost if you pay for them like an honest person. Pretty cheap, right? I try to keep the prices down because I know lots of readers don’t have much money. And anyway, readers aren’t using my books for profit – they’re just reading them.
2. For the years I spent learning to write; for the hard-won skill and knowledge I put into writing those books, for choosing which words came next, and where the commas should fall, I’ll charge $1,000,000.00. Pretty cheap, right? I know, it should be more, because I’ve spent decades learning how to write. But I’m trying to keep the price down because I’m a reasonable person.
3. So, that’s a total of $1,000,012.00
4. Now we know the price, we come to the little matter of whether I’m willing to sell. I’m not so sure I am willing to sell my books and my skills to Meta or to any other tech company. But they can come to me, sure, and ask. Who knows, I might ponder the matter for a whole five seconds. Or I might just get on with writing the next book.

So, if your business uses AI, you might consider asking the company that developed the tool where it got its dataset for training. Did it pay for those books and articles? Did it ask permission from authors to use their skill in such a way?

If you get answers at all, I think we can guess what they’ll be.

But until we know for sure, that’s why I don’t believe a business with any integrity can continue to use generative AI tools for writing.

I’ve sent a letter via the Authors Guild demanding that AI companies remove my works from their training data. If you’re an author and would like to do the same you can use the Authors Guild’s letter template to send your own:
https://actionnetwork.org/letters/authors-guild-author-letters-to-ai-companies?source=email&