Why AI Companies Are Shredding Physical Books

Tech companies training artificial intelligence models are finding unusual ways to source training data. A service provider named ISBNdb is currently marketing physical book orders of up to one million titles directly to developers. The catalog includes rare and out of print materials that have not yet been digitized.

This trend highlights a strange quirk in recent copyright decisions. A 2025 court ruling found that converting a lawfully purchased print book into a digital copy counts as fair use if the physical original is discarded. Because the digital scan simply replaces the purchased book without increasing total copy counts, destroying the paper version creates a clean legal argument for tech firms.

Major players have already engaged in similar practices. Previous court records showed that Anthropic purchased millions of print books, scanned them, and then discarded the paper originals to train large language models. While vendors emphasize secure recycling and compliance, critics worry about the physical destruction of cultural artifacts.

For the crypto and tech markets, legal battles over artificial intelligence data sourcing will likely shape future regulatory standards. Observers should watch for upcoming court decisions regarding copyright law and fair use, as these rulings will directly impact how tech giants acquire their training materials.