ISBNdb, a company that catalogs book metadata, is now marketing physical-book orders of up to one million titles to AI developers. The offering includes non-circulating, rare, and out-of-print material — a move that lands in a legally gray area after a recent copyright ruling.
The June 2025 Ruling
A federal copyright decision on June 23, 2025, held that converting lawfully purchased print books into non-distributed digital library copies is fair use — but only if the physical copies are discarded, keeping the library's copy count unchanged. The same ruling found that Anthropic's use of those copies to train specific large language models was also transformative fair use. However, the court denied Anthropic summary judgment on its pirated central-library copies, leaving those claims for trial. The ruling does not grant a general license to acquire books by any means, nor does it treat every destructive scan as lawful.
Anthropic's Scanning Program
Anthropic hired Tom Turvey in February 2024 to develop a lawful route to a larger research library. The company then spent millions buying print books, having them scanned, and discarding the physical copies. That program is separate from ISBNdb's current service — no common buyer or completed engagement for ISBNdb has been identified.
ISBNdb's Current Offerings
ISBNdb promotes destructive scanning followed by verifiable destruction or recycling. It requires a legally binding nondisclosure agreement for each engagement. The company claims pre-2022 print books are less exposed to AI-generated text and data-poisoning techniques, making them attractive as training data. No evidence exists that a rare, unique, or last-surviving book or edition has been destroyed by Anthropic or an ISBNdb customer.
The HedgieMarkets Post and Unsubstantiated Claims
A July 2026 HedgieMarkets post fused Anthropic's historical project, the 2025 ruling, and ISBNdb's 2026 marketing into one account. Its claims are not substantiated. The post appears to conflate separate source chains — ISBNdb's current service and Anthropic's historical scanning program — with no evidence linking them.
The legal landscape for AI training on physical books remains unsettled. The June ruling left key questions open, and no court has yet addressed ISBNdb's specific business model. For now, AI developers weighing bulk book orders will have to navigate a patchwork of copyright law — and hope the next ruling doesn't come with a cease-and-desist.
