Australian booksellers say an unusual surge in bulk purchases of rare and out-of-print books has fuelled fears that the titles are being bought for destructive scanning to train artificial intelligence models.
The concerns follow disclosures in a US copyright case showing that some AI companies cut apart physical books to digitise them before discarding the originals.
Booksellers warn the practice could threaten literary heritage by removing rare physical works from circulation, while AI companies say they digitise legally acquired books and do not target rare or antiquarian collections.
Concerns are growing among Australian second-hand booksellers over the possibility that rare and out-of-print books are being purchased in bulk for destructive scanning to train artificial intelligence models. Booksellers fear the practice could permanently remove valuable physical works from circulation, particularly as demand for older and hard-to-find titles appears to be increasing.
According to The Guardian, several Australian booksellers have linked these concerns to an apparent rise in bulk orders from buyers they suspect may be part of the AI supply chain. It added that booksellers described the practice of buying books, cutting off their spines to scan each page and discarding the originals as "horrific", even though they acknowledged it is not a new phenomenon.
Why Are Australian Booksellers Raising The Alarm?
The booksellers noticed unusual trend of purchasing patterns. The Australian secondhand booksellers have been wondering if they have been caught up in AI supply cycles as multiple vendors said they received orders from a Canadian company called Zoom Books, which describes itself as a book recycler.
_1642030086.jpg?auto=format%2Ccompress&fit=max&format=webp&w=768&dpr=1.0)
Notably, the orders usually came through large secondhand marketplaces like Abebooks or Biblio, where inventory can range from cheap paperbacks to rare, expensive items, The Guardian report added.
Delfina Manor, who runs Good Reading Secondhand Books in Benalla, Victoria, said she received orders from Zoom Books in May. The order was huge- filling up three full boxes. So she wrote to the company and asked them not to order in such volume again.
“They ordered, paid in advance, and they didn’t quibble over the postage … [but] it would have been about 30 to 40 books, and so finding them, packing them, making sure you hadn’t missed one out … It was just driving me nuts,” Manor said, as quoted by Guardian.
How Does ‘Destructive Scanning’ Work?
These fears about ‘destructive scanning’ has booksellers on their toes as the US- based AI Company Anthropic was entangled in a lawsuit by book authors.
Earlier this year it was revealed that the company was engaging in the practice of “destructive scanning” – buying physical books, slicing off the spines to more efficiently scan the pages, then pulping the remains.
The judge ruled this was lawful under US copyright law, as it amounted to “transformative use”.
According to reports, this process is widespread among AI companies, and now intermediaries are mushrooming who help pace up the process.
Why Are Rare Books Different?
Unlike mass-market paperbacks, rare and out-of-print books are often no longer in print, making them difficult or even impossible to replace once they leave circulation. Many also carry historical, cultural or research value beyond their textual content, making concerns over their destruction particularly acute among collectors, libraries and independent booksellers.
Booksellers say the concern is not simply about the volume of books being purchased, but the seemingly random mix of niche and decades-old titles attracting buyers. Such books may have limited commercial demand but remain valuable because of their scarcity or specialised subject matter.
John Sainsbury, owner of Melbourne's Sainsburys Books, told The Guardian that his store had received a series of orders from Zoom Books for unrelated titles spanning subjects including engineering, cycling, poetry and local history. He said the orders appeared price-insensitive and unusually varied, fuelling speculation that the books were being acquired for large-scale digitisation rather than conventional resale.
Booksellers argued that the debate is not just about digitising books, but about preserving physical copies that carry historical, cultural and archival value. They say destroying one-of-a-kind or out-of-print titles to create digital datasets risks an irreversible loss to literary heritage.
What Are AI Companies Saying?
AI companies maintain that acquiring printed books remains a common way of building training datasets, but reject suggestions that rare or antiquarian works are being deliberately destroyed for that purpose. They say their sourcing practices rely on legally acquired books from commercial markets and distinguish between ordinary second-hand titles and rare collections.
A Spokesperson from Anthropic told the newspaper that they never bought from Zoom Books. But “sourcing books is a widely used approach for training large language models across the AI industry. None of our data acquisition programs buy and destroy rare or antiquarian books.” Anthropic procures its books from mainstream commercial markets, the spokesperson said.
The scrutiny has also turned to intermediaries that buy and resell second-hand books, as booksellers seek greater clarity on where bulk purchases ultimately end up. While some sellers suspect the books may be entering AI supply chains, companies involved in the trade dispute that characterisation.
Zoom Books Co-Founder, Manroop Gill, said in a post on LinkedIn that they are “expanding aggressively”. A spokesperson from Zoom Books told The Guardian, “We acquire secondhand books and resell them intact. We do not digitize books. Our priority is always reuse, and when a book can no longer be rehomed, it is responsibly recycled.”
As AI companies race to secure more training data, the debate is expanding beyond copyright to questions of preservation and cultural heritage. For many booksellers, the concern is no longer just what AI learns from books, but whether irreplaceable physical copies could disappear in the process.
