US AI firm Anthropic scanned and destroyed millions of books after legal purchase
A federal judge ruled the process lawful, while London rare book dealers report anonymous bulk enquiries
Trust 68Craft 77Hype 55How this was reported ▾
Court documents cited and multiple outlets referenced support the claim.
How well corroborated and evidenced the reporting is. Higher is better.
Explains significance and provides specific figures, but lacks direct quotes from affected parties.
Context, balance and separation of fact from comment. Higher is better.
Headline and language use dystopian framing, adding sensational tone beyond the facts.
How far presentation runs ahead of substance. Lower is better.
1 source assessed · methodology
Anthropic, a US artificial intelligence company, bought, scanned and destroyed millions of physical books as part of a project to train its AI models. A federal judge in the United States ruled on 29 July 2024 that the process was protected by fair use. The company legally purchased each book, converted it into a digital copy and destroyed the original.
The ruling followed a £1.13bn settlement Anthropic agreed to over claims it had previously used pirated digital books. Futurism reported that internal documents showed the company launched an initiative called 'Project Panama' after concluding books were essential for training advanced AI models. Executives sought to keep the project private, stating they did not want it to become public knowledge.
Waltham Forest approves £11m Chingford crematorium despite capacity concerns
Council says facility is needed for residents but critics question location and financial viability
London rare book sellers report anonymous bulk orders
Rare book sellers in London have received unusually large enquiries from anonymous buyers. One London-based seller told City AM they were contacted twice by email addresses with no names or company details attached. Both buyers requested additional photographs of specialist books before placing sizeable orders. When asked for their identities, neither responded.
The enquiries match reports from antiquarian booksellers across Europe. 404 Media reported a sharp rise in bulk purchases of obscure titles, with buyers typically concealing their identities. Social media users have described the practice as 'dystopian'. Suppliers are now sourcing between 1,000 and one million books at a time for AI developers, according to 404 Media.
England and Wales record driest July since 1911, Met Office confirms
Provisional data shows rainfall in England was less than half the previous low, with half the country now in drought
Why older printed books are prized for AI training
AI companies prefer printed books published before 2022 because they are less likely to contain AI-generated content. ISBNdb, a company offering bulk book-buying services to AI labs, calls printed books 'the world’s best AI training data'. It requires strict non-disclosure agreements to protect the identity of buyers, acknowledging the reputational risks of the practice.
Anthropic’s internal documents, as reported by Futurism, described books as a way to teach systems 'how to write well', rather than relying on 'low-quality internet speak'. To build its dataset, the company bought millions of second-hand books, sliced off their spines using industrial cutting machines, scanned every page and recycled the rest.
What happens next
The federal court ruling establishes a legal precedent for the destruction of physical books after scanning, provided the originals are legally purchased. It is unclear whether other AI companies will adopt similar practices. Rare book sellers in London and across Europe report no formal mechanism to verify the identities or intentions of bulk buyers.
Questions this report answers
+Why did Anthropic destroy millions of books?
Anthropic bought, scanned and destroyed millions of physical books to train its AI models. A US federal judge ruled the process lawful under fair use because the company legally purchased each book and destroyed the original after scanning. The company aimed to avoid AI-generated content in its training data.
+How much did Anthropic pay in its settlement?
Anthropic agreed to a £1.13bn settlement over claims it had previously used pirated digital books. The settlement was reported by Futurism and relates to the company’s earlier practices, not the physical book scanning project.
+Are London rare book shops being targeted by AI firms?
London rare book sellers report anonymous bulk enquiries for specialist books, with buyers refusing to disclose their identities. One seller told City AM they received two large orders from email addresses with no company details. Similar reports have emerged across Europe.
+What happens after the US court ruling?
The US federal court ruling establishes a legal precedent for the destruction of physical books after scanning, provided the originals are legally purchased. It is unclear whether other AI companies will adopt the practice. Rare book sellers have no formal way to verify the identities of bulk buyers.
WE FILE
No invented reporter. No hidden source trail.
The HNL Local Desk assembled this report from the outlets listed below. Language-model drafting is reviewed under human editorial supervision, with scoring notes and corrections kept visible.
Harrow AM
The compact morning rundown, with scores and source links.
More Coverage in World
Waltham Forest approves £11m Chingford crematorium despite capacity concerns
England and Wales record driest July since 1911, Met Office confirms
Jay Archibald jailed for three years and nine months for Stratford Overground robbery
