European Writers' Council denounces AI firms' destructive scanning of millions of books

What happened

On 25 September 2026, the European Writers' Council (EWC) published a statement denouncing what it calls the "destructive scanning" of books by AI companies: the bulk purchase of at least two million old, rare, and out-of-commerce books through second-hand and antiquarian booksellers, followed by cutting the spines, high-speed scanning, and shredding of the physical copies — with the digitised files used to train large language models. The EWC states the digitised works feed "foundations of e.g. Anthropic's 'Claude Sonnet' or Amazon's AI software," and specifically attributes a secret "Project Panama" — described as aiming to "scan all the books in the world" — to Anthropic.

The statement quotes the 16 September 2026 written answer of European Commission Executive Vice-President Henna Virkkunen to MEP Andreas Schwab (E-002819/2026): the digitisation of copyright-protected books, including lawfully acquired analogue copies, is an act of reproduction requiring the rightsholder's prior authorisation, and the EU text-and-data-mining exceptions apply only to text and data in digital form — not to the digitisation of physical works. The Commission's answer itself was not independently retrieved; this item rests on the EWC's own statement.

Why it may matter to you

If you are an author, this is the first formal EU-level confirmation — via the Commission's answer — that scanning physical books for AI training cannot shelter behind the text-and-data-mining exception. The EWC is calling on the Commission, Parliament, and member states to act, which could produce enforcement or legislation affecting how training corpora are sourced.

What is not yet known

Whether the Commission will open investigations or propose measures beyond the written answer; whether the companies named have licences for the scanned works; and how the quoted legal position interacts with the member states' implementation of the TDM exceptions.