ISBNdb Removes AI Landing Page Amid Backlash Over Destructive Book Scanning
Book database ISBNdb has taken down an AI-related marketing page following public outrage over reports that AI companies are purchasing and destroying physical books to digitize them for LLM training. The platform clarified that it does not train AI models itself and that the page was merely a market demand test. This controversy highlights the aggressive and ethically questionable data acquisition strategies of AI companies seeking high-quality, "clean" training data to avoid model collapse caused by AI-generated web content. It also underscores the legal and ethical loopholes in copyright law regarding the physical destruction and digitization of copyrighted works. The practice of "destructive scanning"—slicing off book spines for high-speed, low-cost digitization—was previously exposed in Anthropic's secret "Project Panama," which a federal judge ruled was legal. ISBNdb's deleted page had pitched physical books as containing structured, peer-reviewed knowledge that web crawlers cannot replicate.
## BACKGROUND
ISBNdb is a comprehensive online bibliographic database that aggregates metadata for books worldwide, including titles, authors, and publishers. As AI companies exhaust easily accessible web data, they face the risk of "model collapse" from training on AI-generated content, driving them to seek high-quality offline sources like published books. Anthropic's "Project Panama" is a prominent example where millions of used books were purchased, had their spines sliced off, and were scanned to train the Claude chatbot.