Not too many years ago, before the dawn of Kindles, used bookstores were a refuge for budget-minded, voracious readers. As the industry evolved, many suffered. Now they’re seeing a resurgence in sales, but the customers increasingly aren’t bookworms. They’re AI companies.
Barter Books, for instance, typically sells 2,000 to 3,000 books per week, which isn’t bad for a used bookstore in Northumberland, the least densely populated county in England. Recently, however, the owner got a bulk order from a company in Canada for that many books in a single day.
Several other bookshops in the area are seeing similar demand. And while they can’t definitively say what the books are being used for, a recent court ruling offers a clue. The fate of those books may also be a dire one.
AI companies have been in hot water with the literary world in recent months. Anthropic recently reached one of the largest copyright infringement settlements in history, agreeing to pay $1.5 billion to more than 300,000 writers who filed a complaint against the AI giant two years ago. OpenAI, meanwhile, is still facing several copyright infringement lawsuits from a variety of sources, including The New York Times, Encyclopedia Britannica, comedian Sarah Silverman, and several nonfiction authors, who accuse the company of copying their books to train its large language model and saying OpenAI has “enjoyed enormous financial gain from their exploitation of copyrighted material.”
The Anthropic suit, however, may have provided a path forward for AI companies. The court ruled that it was not illegal for Anthropic to train its AI on copyrighted works, as long as it paid for the books it used. A separate judge ruled similarly last year for Meta.
Used books offer AI companies a cheaper way to acquire source material. Some publishers might also refuse to supply copies directly to an AI company if they know what it plans to do with them.
The Anthropic suit highlighted a program at the company called “Project Panama,” which Anthropic described as its “effort to destructively scan all the books in the world.” A memo highlighted as an exhibit in the trial explained that the company used a codename for the training method “because we don’t want it to be known that we are working on this.”
Destructive scanning physically disassembles bound books so their individual pages can be fed through high-speed scanners and ingested by a large language model. Once the scanning process is complete, the pages are discarded, recycled, or pulped.
While it may not be an especially big deal if a copy of a mass-market paperback gets destroyed, something that happens all the time, book enthusiasts worry that rare books, which are often sold by used bookstores to collectors, could be swept up in the process as well.
Anthropic tells the BBC that’s not the case, saying “none of our data acquisition programs buy and destroy rare or antiquarian books” in a statement.
It’s unlikely, however, that Anthropic is the only AI company using this method to train its systems, and other companies may not have the same limitations on what they destructively scan. Antique and rare books, after all, often contain information that other books don’t, which could make them particularly valuable as training material.
One recent shipment containing at least one rare book, for instance, was tracked to an Amazon AI training facility in Las Vegas, which destructively scans books for AI training.
Most used booksellers can’t say definitively who is buying books from them in bulk. The orders come through marketplaces like Biblio, which anonymizes buyers. But when thousands of books are ordered at once, it’s unlikely they’re all going to avid readers adding to their personal collections.
While some store owners might have ethical reservations about where the books are going, the cash infusion is welcome. Bookstore sales peaked in 2007 at $17.18 billion and have largely been falling ever since. Although the industry has seen a slight comeback in the past three years, many bookstores have still had to reevaluate their business models or close their doors for good.