The AI company Anthropic is facing a lawsuit for using external content to train its own AI Claude. This involves, among other things, millions of books that Anthropic claims to have purchased.
On Monday, June 23, 2025, court documents revealed that the AI company Anthropic spent millions of dollars scanning printed books to develop Claude, an AI assistant similar to ChatGPT.
One particular detail: The company bought used printed books, cut millions of printed books from their bindings, scanned them into digital files, and discarded the originals, just to train the AI. This was reported by the English-language magazine ArsTechnica.
Company purchases printed books to circumvent licensing difficulties
Why is Anthropic doing this? The problem is generally about the legal control of content. Companies and publishers fundamentally believe that digital content belongs to them and that AI should not use it without payment. And AI companies are not always willing to negotiate a license for it.
The way out of the dilemma: Anyone who buys a physical book can do more or less what they want with it and can also destroy the book they purchased. And exactly this advantage is what Anthropic exploited.
Because purchasing used physical books completely circumvented the licensing issue and simultaneously provided the high-quality, professionally edited text that AI models need, and destructive scanning was simply the fastest way to digitize millions of volumes.
The company reportedly spent “many millions of dollars” on purchasing and scanning, often buying used books in bulk. The books were then removed from their bindings, the pages trimmed to a usable size, scanned as stacks of pages into PDF files with machine-readable text including the covers, and then all paper originals were discarded.
What was the judge’s decision? In favor of the AI company: Judge William Alsup stated that this destructive scanning process should be considered fair use. However, only because Anthropic legally acquired the books first, destroyed each printed copy after scanning, and kept the digital files internally rather than distributing them. Had the company distributed the scans, there could have been legal issues.
The judge compared the procedure to “preservation of space.” However, he did not elaborate much further.
A company must file for bankruptcy. For years, it had offered AI services that were actually performed by humans. Companies like Microsoft are said to have invested millions in the company: A company pretended to work with AI for years: Exposed for secretly employing 700 real people
