Secondhand Book Sales Rise as AI Firms Buy Books for Training Data

Booksellers report unusual bulk orders for old books, and evidence suggests AI companies may be buying them to train chatbots.

maisiekooc
Maisie Morrison

AgentLocker Editor

AI News
Secondhand Book Sales Rise as AI Firms Buy Books for Training Data

Booksellers around the world are noticing something odd. Large bulk orders for secondhand books are being shipped out to distant warehouses, and nobody is quite sure why.

Stuart Manley runs Barter Books in Northumberland. He normally sells two or three thousand books a week. Recently, a single order from a Canadian company matched an entire week's worth of sales.

He says he has never seen anything like it in thirty years selling used books.

Other booksellers report similar experiences. Large, unexplained orders have become common, and nobody can say for certain where the books are ending up.

The working theory is that these books are not going to readers. Instead, they may be feeding something else that needs huge amounts of information: artificial intelligence.

The Court Case Behind the Boom

The theory traces back to a US court ruling made in 2025. A judge decided that using purchased books to train AI software did not break copyright law.

The case involved three authors who sued the AI company Anthropic. Judge William Alsup ruled that Anthropic's use of the books was "exceedingly transformative" and allowed under US law.

Court documents unsealed last month showed that books were being destroyed during the process of training Anthropic's chatbot, Claude.

A spokesperson for the company said Claude is trained using a mix of public web data, purchased datasets, and material the company creates itself. They said sourcing books this way is common across the AI industry.

They also said none of the company's data programs buy and destroy rare or antique books.

What "Destructive Scanning" Means

Court papers referred to the book buying project internally as "Project Panama." The stated goal was to scan every book in the world.

This process is known as destructive scanning. Workers remove the spine of a book so every page can be scanned quickly, then the remains are recycled.

Manley says he had not heard the name Project Panama before, but the practice itself is already a hot topic on bookseller forums.

He does not know if his own stock has gone toward this specific project. He says the orders seem random, covering everything from obscure Latin texts to old cowboy novels.

Experts say that mix makes sense for AI training. Rare or unusual texts can give language models fresh material they have not seen before.

Professor Emily Hudson, an intellectual property expert at Oxford University, points out that UK copyright law works differently than US law. She says copying books to build a training library generally requires permission from the copyright owner in the UK.

For booksellers, the situation creates a dilemma. They welcome the sales but worry about what happens to the books afterward.

Derek Walker, who owns McNaughtan's in Edinburgh, says some books are common enough that losing one copy is not a major loss. But he says his shop has also sold books believed to be the only surviving copy of an 18th century edition.

He says destroying a book like that would be a real loss.

Manley takes a more practical view. He says recycling makes sense for books nobody wants anymore, especially when it brings in business.

"The world no longer needs five million copies of The Da Vinci Code," he says. "I've had books advertised for 20 years on the web which haven't sold until now."

From our research desk
AI Jobs Automation Index
Which jobs are AI tools targeting most? We mapped 3,400+ AI tools to real job functions — with BLS employment & salary data.
Explore the index
maisiekooc

Written by

Maisie is a news writer at Agent Locker, covering the latest developments in artificial intelligence, emerging technology and the companies shaping the future.

Discover AI Agents