Talk about insult piling onto injury . . . AI is inflicting both on authors these days.
Yes, AI is gushing out books, creating a flood threatening to drown all books and authors in its path, while also taking out unsuspecting readers.
AI tech bros stole published works despite copyright and used it to train their LLMs (Large Language Models, which is a more accurate description of what is being called AI.)
And now there are apps and programs claiming they can spot AI works, but there are serious, serious issues with that.
Let’s take those one by one.
AI is gushing out books
The stats being tossed around are astonishing and numbing. They’re also estimates. Nobody truly knows how many books being released are AI “written.”
In fact, no one knows how many books have been released in recent years — period. Many of the so-called stats cover only traditional publishing, ignore every book released by every self-published author. Other estimates only include those who pay for ISBNs (International Standard Book Number.) I’m among the U.S. authors who pay for the privilege of having ISBNs. (UK and Australian authors/publishers also have to pay for ISBNs. Canadian and New Zealand authors/publishers don’t.)
Still, within the clunky and incomplete estimates, the increases are eye-popping.
“Commonly cited estimates” for ISBN-registered books in the U.S. only are:
2.0 million in 2022
2.3 million in 2023
3.1 million in 2024
4.1 million in 2025
Ignoring the specifics of what those commonly cited estimates include or miss, it’s worthwhile to look at the jump as AI has spread. Want to take a guess what 2026 will be?
AI tech bros stole published works despite copyright and used it to train their LLMs
I’m among the authors who have works covered in the class action lawsuit settlement with Anthropic.
That settlement says they used 17 of my books to train the Anthropic LLMs.
Let me rephrase that. Anthropic acknowledges using 17 of my books to train its LLMs after obtaining those books from pirate sites. In other words, they downloaded and benefitted from stolen works.
Anthropic also used books it bought. For those, the authors get no compensation. Nada, zilch, bupkus. The judge said, copyright-schmockyright, it’s a-okay, go right ahead, grab the copyrighted material that took the author months — not to mention creativity and discipline — to write before, in the case of self-published authors, then paying for editing, a cover, formatting, marketing, copyrighting, uploading. But, hey, tech bro, as long as you paid $5 or $6, maybe even — gasp — $7 or $8, you’re fully entitled to use it in its entirety however you want to benefit your multibillion dollar enterprise.
In addition, I am certain that at least three more of my books were used (backed by logic, including the inclusion of the copyright number for them being included in the settlement listing), but the “rules” of the settlement say tough luck. Oh, wait, they also say I can find my own lawyer, pay him or her, and sue Anthropic over those three titles, along with all my other books that were published before the date of infringement. Because, guess what? All my books have been pirated.
Every single one of them.
Most of the pirate sites are scams that grab would-be users’ info and give them nothing but well-deserved headaches. But a few, apparently, were safe enough for Anthropic to risk lifting books from — admitting to about half a million such books.
To be clear, Anthropic is absolutely not the only AI company that did this. It is the only one — so far — that has settled, giving the authors whose creative works are the backbone of its product any compensation at all. Well, promising to give us a measure of compensation. Haven’t received any yet.
And now there are apps and programs claiming they can spot AI works, but there are serious, serious issues.
Stick with me here. The LLMs were trained on my works, on the works of so many other authors. Our voices, our word choices, our characters, our plots, our punctuation (we’ll come back to that), our everything.
Now authors are running books written and published well before AI existed through Oh-Great-And-Powerful apps and programs, which are then reporting they’re 80%, 90%, even 100% certain the book in question was AI-generated. Let me repeat that — the apps and programs are labeling books written and published before AI existed as AI generated.
With the potential to do great damage to an author. (Here is one writer’s experience running new writing through Pangram, supposedly the best of these apps/programs, and changing/editing to try to win its approval.)
Huh? Nobody creating these apps and programs realized that what they’re “testing” might seem like AI to them because AI is based on exactly that original material — the material stolen to train the LLMs?
LLMs are predictors of the next most likely word. Its predictions are based on what was learned from ingesting all those original works.
These works don’t resemble AI, AI is a ripoff of these works.
Sheesh.
And that brings us to punctuation. Specifically, the em dash, which is the long dash I’ve used several times in this piece. (For those interested, there’s the hyphen, the en dash, and the em dash — in order of increasing length. En and em are long-time typographical measurements, says the long-time newspaper journalist.)
According to these oh-so-marvelous-would-be-AI-spotter apps and program, the use of em dashes is a sign of AI-generated writing.
That smoke you smell is my hair on fire.
Look, bub, I’ve been using em dashes for rhythm and emphasis since before my first book was published in 1990 — yes, all those books AI stole to train the LLMs had bunches of em dashes in them. Heck, that might be why these idiot apps and programs associate em dashes with AI — because the AI tech bros stole my em-dash-laden writings to train their AI spewers.
~ ~ ~
To my readers, as I’ve said before, I will tell you if I start using AI as more than a modestly improved search function to check what I named the guy at the hardware store without flipping through eight or eleven or fifteen books.
For now, I thank heavens I write murder mysteries, because I have a plethora of candidates worthy of killing off—
Cue the em dash.
—fictionally.