Even Elon Musk opposes AI firms ‘shredding’ rare books to train LLMs — now, the world’s largest book database denies aiding them

“Once a final non-digitized copy of a book is shredded, it’s gone forever,” an editorial in Scotland’s Herald warned. “And the version that is digitized becomes at the mercy of AI interpretation rather than a straight, accurate representation of the work.”

The practice first surfaced last summer when a lawsuit against Anthropic uncovered internal documents referring to “Project Panama“ as, allegedly, an “effort to destructively scan all the books in the world.” A recent story about the issue by 404 Media re-ignited reader rage with descriptions of machines slicing off book spines to “feed” the AI scanners.

Even Elon Musk came out publicly to oppose the practice.

“I’ve asked the SpaceX AI team to preserve any rare books in a library and scan them the hard way vs. just cutting off the spine and scanning,” Musk posted to X on July 27.

It’s not just AI companies taking the heat, however. ISBNdb, which dubs itself “the world’s largest book database,” was specifically singled out in the 404 story for a since-deleted article on its website that appeared to condone the AI book-butchering practice — putting them at the center of the firestorm around how AI companies may use and abuse humanity’s literary legacy.

The surprising value of old books in the AI race

The Washington Post reported that Anthropic favored print books because they’re well written and free of the “low quality internet speak“ that often fills the web.

Meanwhile, the battle over Anthropic’s use of books to train its LLMs — including those that power Claude — hit the courts last summer. Part of it, as the Washington Post noted, centered around Anthropic’s alleged use of downloaded, “pirated” books without author permission.

A judge ruled that using pirated books was illegal, but digitizing legally-purchased print books qualified as fair use and “transformative” because they weren’t reproduced or redistributed but, rather, became searchable digital copies in Anthropic’s central library.

As such, Anthropic settled with the authors they allegedly pirated for $1.5 billion while being given a pass for what the Washington Post has described as “[spending] tens of millions of dollars to acquire and slice the spines off millions of books, before scanning their pages to feed more knowledge.”

Fast-forward to this July, when the 404 story highlighted the issue while pointing to a post on ISBNdb’s website that reportedly touted that “the world’s best AI training data is sitting on a shelf.” ISBNdb also reportedly claimed that “it can keep the identity of AI companies secret” if they purchase books and that “the optics problem is real,” adding that “‘AI company destroys two million books’ is not a headline that generates sympathy.”

Those posts are no longer available on the ISBNdb website and a company rep reiterated to Moneywise that they’ve “never bought, sold, or scanned printed books for AI training.”

They added that the “related content” on their site “was written for the service [they] were exploring, and once it was clear the pages were being read as describing an active service, [they] took them down.”

As for whether that service could still become active, ISBNdb said they “don’t comment on product roadmap” but that “no such service was ever brought to life, and ISBNdb isn’t offering it.”

A global book-buying spree is raising eyebrows

Independent or secondhand booksellers around the world, meanwhile, are reportedly fielding requests for large troves of old or rare books that many believe are coming from AI companies.

One rare book seller told 404 Media that, in recent months he and other colleagues “noticed a historic spike in sales.” He explained that he usually only sells about 20 books a week and that, while he doesn’t like that “uncommon books are being pulped,” it “benefits [him] financially … by clearing out old inventory that is otherwise unlikely to sell.”