Why Anthropic's Book-Scanning Practice Draws Scrutiny

In one line: The Guardian looks at Anthropic's practice of scanning physical books and then discarding them to build training data, unpacking the context and the controversy.

Key points

• As AI firms need vast amounts of text to train large language models, digitizing physical books by cutting and scanning them has come under scrutiny.
• Scanning reportedly requires slicing off each book's binding, effectively destroying the original — a detail that has fueled symbolic backlash.
• The practice is tied to ongoing debates over copyright and fair use around digitizing lawfully purchased books.

Why it matters

With the provenance and acquisition methods of AI training data now central to copyright litigation and regulation, physically discarding books highlights where technical justification collides with cultural sentiment. Transparency in how training data is sourced is likely to remain contested.

Read more

• Why is Anthropic destroying books? — The Guardian