When Books Become Training Data: Anthropics's Project and Why Provenance Still Matters

In 2024, Anthropic (parent company to Claude AI) ran an internal project called Project Panama. The planning document described it plainly: “Project Panama is our effort to destructively scan all the books in the world.” The same document noted: “We don’t want it to be known that we are working on this.” That document was unsealed in federal court.

I’ve been reading since before first grade. I’m not exaggerating — I was the kid with a book at the dinner table. I never stopped. Right now there are books stacked on my nightstand, books on my office shelf, books I return to. My husband Jay has books that belonged to his grandfather. Those books aren’t going anywhere. They have provenance. They have weight. They’re the kind of thing you don’t just recycle or sell.

When I read about Project Panama and shared with Jay,  we both felt it. 

What They Actually Did

Anthropic (parent company to Claude AI) purchased used books in bulk through platforms like Better World Books and World of Books. They used hydraulic spine-cutting machines and high-speed industrial scanners. After scanning, the paper was recycled or pulped. The target was 500,000 to 2 million books in a six-month window. Cost: tens of millions of dollars.

One copy purchased. One copy destroyed. One digital file kept.

A federal judge, William Alsup, ruled that the scanning of legally purchased books was “exceedingly transformative” fair use. The logic: Anthropic bought the book, Anthropic destroyed it, Anthropic kept one digital file internally. No public distribution. Legal.

That ruling covers the books they bought and shredded.

The Settlement That Just Landed

Eight days ago — July 20, 2026 — a federal court gave final approval to a $1.5 billion settlement between Anthropic and authors. This was a separate matter entirely. These were pirated books, downloaded from LibGen and similar repositories. Books that authors never licensed, never sold to Anthropic, never consented to include in any training set.

$1.5 billion. Final approval eight days ago.

Not a footnote. It’s the current news.

The Cultural Cost

Physical books are stubborn. You can’t push an update to them. You can’t alter what they say after the fact. You can’t put them behind a paywall once someone owns a copy. That stubbornness, it turns out, was treated as a problem to be solved.

The preference in bulk book acquisition — and this isn’t just Anthropic; brokers like ISBNdb facilitate anonymous bulk orders of up to a million books. NDAs and coaching to call it “digital preservation” — is for pre-2022 books. Older stock. Human-curated text. No AI contamination.

The books most valuable for training are the ones that represent the clearest record of how human beings actually thought and wrote. And the process of extracting that record destroys the object.

Carnegie Built Libraries. Anthropic Built a Shredder.

On July 27, Brian Roemmele put it starkly on X: “The robber barons built cathedrals of knowledge. Anthropic built a shredder.” Andrew Carnegie funded roughly 2,500 public libraries. His wealth, extracted from labor, was channeled back into public access to books.

Whatever you think of that history, the contrast is hard to ignore.

On the same day, Elon Musk — whose xAI, now operating as SpaceXAI(aka Grok) under SpaceX, was already doing large-scale book scanning — replied on X that he’d asked the SpaceX AI team to preserve rare books in a library and scan them “the hard way vs just cutting off the spine and scanning.” It’s a different instinct — that the physical object has value beyond the text it contains.

A Bookseller’s Take

Not everyone reads this story the same way. Charlie Becker, a second-generation bookseller whose family runs Houston’s largest used and rare bookstore, posted on X after his own store was hit by bulk orders. His read: the ingredients of the viral story are all true, but “rare” often means a 1995 Denver travel guide or a WordPerfect manual — obscure, but not precious. His concern was quieter and darker. Books that don’t sell through Amazon’s FBA system eventually get liquidated. Nobody has to shred them. They just disappear. His full writeup is worth reading.

His closing line landed with me: “The Library of Alexandria is still burning, just more slowly, and more from ‘who cares’ than from some cinematic villain.”

The Library of Alexandria, as imagined by a 19th-century artist. Public domain.

Why I Built Quoteiac the Way I Did

At Quoteiac, we have one rule that governs everything: if we can’t verify it, we don’t print it.

That rule came from the same place this story hits me. Attribution matters. The source matters. Knowing where something came from — who wrote it, when, in what context — matters. A quote stripped of that is just words. Anyone could have said it. No one owns it. It floats.

Before I started this business, I kept running into misattributed quotes. Famous names attached to lines they never wrote. I couldn’t do that. Every piece we make traces back to a verified source. That’s our brand promise — and the way I think about language and ownership and the record of human thought.

Project Panama is now in the light. The text gets extracted. The physical object — the margins, the previous owner’s underlining, the binding that held it together — gets destroyed. What remains is a file. Clean, searchable, ownerless.

Jefferson Knew

Thomas Jefferson wrote: “I cannot live without books.” He wrote it in 1815, after selling his personal library — 6,487 volumes — to Congress to replace the books the British had burned in the War of 1812. Then he immediately started building another library. He couldn’t help it.

Jefferson understood something that is easy to forget: books are not a leisure item. They are infrastructure. The accumulated record of how people thought, argued, made mistakes, and occasionally got something right. He was two hundred years ahead of this moment, and he saw it clearly.

We have a verified Jefferson quote page and a full piece on Jefferson and books in the Journal if you want to go deeper. The quote earns its place here.

What You Can Still Do

I’m not telling anyone to stop using AI. I use it myself. But I do think the physical book deserves a defense, and there are concrete things any reader can do.

• Buy from independent bookstores. Used-book platforms with bulk acquisition programs are the supply chain for this kind of operation. Independent stores aren’t.
• Keep the books you have. Jay’s grandfather’s books are staying on that shelf. Provenance is real. The chain of ownership means something.
• Write in your books. Underline them. Date your notes. Marginalia is not recoverable from a scanned file. It’s the one part of a book’s life that a shredder actually destroys.
• Support authors directly.  Pre-order. Show up to readings. The $1.5B settlement covers some of the harm — but only some, and only retrospectively.
• Be skeptical of “digital preservation.” When a bulk acquisition program uses that phease, ask what happens to the physical copies afterward.
• Pay attention to the legal record. The Bartz v. Anthropic unsealing happened because someone pushed. These documents don’t surface on their own.
• Value old books specifically. Pre-2022 stock, human-curated, with no AI in the text — that’s what the acquisition programs want most. It’s also, not coincidentally, where a lot of the literature is.

The Shelf Holds

Jay’s grandfather’s books are still here. They’re not in a database. They’re not a training file. They’re objects — with a particular weight and smell and the faint possibility that someone underlined something in the right margin sixty years ago.

I think about that when I think about what gets lost when the spine gets cut and the paper gets pulped. Not the text — the text survives. What gets lost is everything else.

At Quoteiac, every word we print traces back to a verified source. That’s the whole thing.

Words with provenance. If we can’t verify it, we don’t print it.

More from the Journal · Thomas Jefferson on Books, Reading, and Ignorance · The Object That Holds the Thought · The Declaration of Independence Is Great Literature

About the Author

Vickie MacFadden is the founder of Quoteiac. She has been reading since before first grade and has never stopped. The provenance of a book — who held it, who wrote in the margins — matters to her. It always has.