How Anthropic's Project Panama destroyed millions of books to train Claude — and won in court
创业邦 · wechat · 2026-09-20
A deep-dive into Anthropic's Project Panama: the company spent tens of millions of dollars buying millions of physical books (mostly second-hand), guillotining their spines and high-speed scanning them to build a permanent research library for training data, with vendor plans to process 500k–2M books in six months. Court filings show scanned books and 7M+ pirated titles (Books3, LibGen, PiLiMi) feed a central library far larger than what models actually train on. Legally, training was ruled transformative fair use; destroying the originals helped prove no duplicate copy was created, while only the piracy led to a $1.5B settlement. A B2B physical-book procurement industry for AI firms is now emerging, favoring pre-2022 books as 'uncontaminated' data.
Related event: Anthropic's Mass Book-Scanning Project for Claude Training Exposed(2 posts)→
More from AGI Musings
- Andreessen: Agents Are Entering Enterprises as Chaos, Just Like PCs and Spreadsheets Did — Dan_Jeffries1 · 2026-11-04
- 1000+ mathematicians sign pledge to shun AI companies and models — Shesawthat · 2026-11-04
- Von Neumann meme: "You don't have to be responsible for the world that you're in" — akbirthko · 2026-11-04
- Joscha Bach on the dream chocolate that stopped being real — Machine Learning Street Talk · 2026-11-04
- François Chollet: modeling brain wiring as an engineering problem constrained by genome size and time — CatAstro_Piyush · 2026-11-04
- Most US Couples Now Meet in a Data Centre, Doubling Office Matches — menhguin · 2026-11-04