Companies are now buying large numbers of physical books, extracting the information they want, and then disposing of the books afterwards. Once the data has been taken, the physical books may be removed from public access, stored without oversight, or destroyed. Entire collections can disappear simply because a corporation no longer needs the physical copies.

This practice has become especially visible in the AI industry. Public reporting has raised concerns about companies such as Anthropic, along with other major AI developers, acquiring books in bulk for model training and dataset expansion. If you want to see the evidence for yourself, here are two publicly available sources:

The Atlantic reported that Anthropic purchased over 100,000 books from a private seller for AI training.
The New York Times has covered how AI companies use copyrighted books to train models without clear compensation frameworks for authors.
These articles show that book acquisition for AI training is already happening — and that transparency about what happens to the physical books is limited.

A recent US court ruling has intensified global concern. The ruling suggested that using copyrighted books for AI training may be allowed under US law. Although this ruling applies only within the United States, it has worldwide consequences: if US companies can legally use books for AI training without clear compensation or preservation requirements, then books from any country could be acquired, digitised, and potentially discarded. This means a decision made under US law could affect books, authors, and cultural heritage everywhere.

This creates a significant power imbalance.  If AI companies can legally acquire and use books without clear international rules, they gain the ability to absorb vast amounts of human knowledge while the public loses access to the physical works themselves. This concentrates cultural and informational power in private hands: companies can keep the data, discard the books, and benefit from the knowledge without preserving the original sources or compensating the authors. Without international protections, a small number of corporations could end up controlling enormous datasets built from books that no longer exist in public libraries or collections.

Books are not disposable raw material. They are part of humanity’s shared cultural memory — the stories, knowledge, and history that shape who we are. When corporations treat books purely as data sources, their cultural, educational, and historical value is lost. Readers lose access, authors lose income, and cultural heritage can quietly disappear into private datasets.

There is currently no international legal framework preventing corporations from acquiring and disposing of books without oversight. No preservation requirements. No transparency rules. No guarantee of copyright payments to authors. Nothing to stop companies from destroying books once they have extracted what they want.

We call on UNESCO, WIPO, and national governments to act now.  We ask for international rules that prevent corporations — including AI companies — from destroying books after data extraction; require full transparency when books are purchased in bulk; mandate the preservation of physical books; and ensure that authors receive fair copyright payments whenever their work is used for digitisation, data extraction, or AI training.

If we do nothing, more books will be lost, more authors will be underpaid, and more of our shared cultural heritage will disappear forever.