The recent uproar over the wholesale trashing of books to train AI models is showing no signs of abating as additional reporting has kept the story alive. 404 Media, which broke the initial story involving an aborted marketing campaign by a data company offering to supply books in bulk for AI training, recently inserted an air tag into what it claimed was a rare book and tracked it to an Amazon-owned facility in Las Vegas, whose sole business is the destructive bulk scanning of books.
The facility, VGT-3, didn’t make itself look any more sympathetic by adopting a logo featuring the image at the top of this post.
A bookseller in the Pacific Northwest reported slipping credit-card sized trackers under the dust jackets of books in two different shipments to suspected AI company buyers and traced one of them to an intersection in Las Vegas with two Amazon facilities and ultimately to a location in Mexicali, Mexico, known to host a large paper recycling center.
But the story has also produced its share of odd bedfellows. Elon Musk, of all people, as culturally backward a boor as you’ll find, asked the SpaceX AI team in a post on X to “preserve any rare books in a library and scan them the hard way vs just cutting off the spine and scanning.” What that’s supposed to accomplish other than producing a nice headline for Musk is unclear.
Second-hand book sellers meanwhile, normally bibliophiles’ allies, are enjoying a windfall from the spike in bulk buying, presumably by AI companies, albeit mostly of titles that won’t be missed. One dealer described a recent large order to the Wall Street Journal as including a textbook about the Russian economy’s transition after the fall of the Soviet Union, a Victorian novel, a 2012 Texas civil procedure guide and a book about a Swedish comedy team popular in the 1960s.
“Cultural artifacts? I can categorically tell you that they’re not.” he said, describing most of the books as dead stock—titles that sat on his shelves, collecting dust for years that he was considering taking to a recycling center himself before the order came in.
But the oddest pillow partner has to be Anna’s Archive. The notorious “shadow library” has been sued repeatedly by book publishers, music companies and other rights holders for, among other things, maintaining a massive collection of digitized texts and other material—compiled mostly without authorization—that is frequently tapped by AI companies for training fodder. Yet, in early August it issued a worldwide appeal for volunteers to scan material from every library and archive on earth and upload the files to Anna.
“As the world’s largest shadow library, Anna’s Archive needs a plan to combat the destruction of physical books by AI companies,” it said in its appeal. “After all, the emergence of shadow libraries is the greatest miracle of knowledge sharing in the 21st century. Along with other shadow libraries, we’re building a digital library of Alexandria, an inextinguishable light of humanity.”
Publishers and authors are unlikely to share the Anna’s enthusiasm for shadow libraries, but they likely share, albeit uneasily, an enthusiasm for keeping books out of the hands of AI companies.
Anna’s appeal raises another issue that also is echoed in unlikely places.
“After AI companies massively scan and destroy physical books, they become the only ones in the world with digital copies,” it said. “Knowledge is permanently monopolized on private servers [sic].” Which it attributes to the depredations of late-stage capitalism.
Behind it lies the AI race and the interests of capital:
It prevents these books from being scanned and used for training by competitors.
It avoids legal risks.
Destroying books is cheaper than lossless scanning.
The Open Markets Institute and the Consumer Federation of America might not be as down on late-stage capitalism, but they were both signatories to a letter to the Federal Trade Commission last week that raised similar concerns. It asked the FTC to investigate “whether and to what extent” the destructive scanning of old and rare books “forecloses competing AI developers and the public from a scarce, non-reproducible input and whether it constitutes an unfair method of competition under Section 5 of the FTC Act.”
Other signatories include the Demand Progress Education Fund, and the Media and Democracy Project, along with a few booksellers.
“To the extent that an AI company acquires and destroys such source materials, then walls off its proprietary training corpora from the world, the company is systematically starving the market,” the letter said. “Unlike a standard data acquisition strategy, this hoard-and-destroy practice could serve as yet another structural mechanism to raise rival companies’ costs and deny start-ups and fledgling competitors a key source material essential to competing in the AI marketplace.”
I wouldn’t expect this FTC to do much investigating given the Trump administration’s pro-AI stance, and certainly not at the behest of Food & Water Watch and the Workers Circle, whose names are also on the letter. But the issue’s odd synchronicities are an example of AI’s ability to scramble traditional political and ideological alignments.
The grassroots opposition to AI data centers, which is coming from both left and right, is the poster-child of the phenomenon. But the two controversies are of a piece. They each reflect opposition to a perceived enclosure of what are viewed as public spaces—whether knowledge spaces, physical spaces, or the psychic space of community—by invasive private interests.
It’s also what distinguishes them from the AI vs. copyright debate. Unlike public domains, copyright is a private interest that is sometimes seen as at odds with the public’s. Copyright industry lobbyists, trade associations and PR firms do their best to cloak themselves in the garments of the everyman artist. But is just not as convincing or persuasive as gazillion-dollar technology companies laying waste to public endowments.