Anthropic Lawsuit Reveals 'Destructive Scanning' of Rare Books

AI Companies Destroying Physical Books After Digitization

Generative AI companies are purchasing physical books—including rare and antique volumes—cutting off their spines to scan pages more efficiently, then destroying what remains, according to court documents unsealed in a copyright lawsuit against Anthropic.

The practice, termed "destructive scanning," has alarmed rare book vendors and preservation experts who fear irreplaceable artifacts are being lost to feed AI training datasets. The Guardian first reported details from the unsealed court files, which revealed Anthropic had sought to keep the practice quiet while engaging in it systematically.

How Destructive Scanning Works

The process involves acquiring physical books, severing the spine to separate pages, and running them through high-speed scanning equipment. Once digitized for AI training purposes, the physical materials are pulped or otherwise disposed of. This approach prioritizes scanning efficiency over preservation of the original artifacts.

For AI companies racing to expand their training datasets, destructive scanning offers speed advantages over traditional preservation-focused digitization methods that keep books intact. The practice appears particularly concerning when applied to rare, antique, or one-of-a-kind volumes that cannot be replaced once destroyed.

Copyright and Preservation Concerns

The revelation emerged through a copyright lawsuit brought by book authors against Anthropic. The case highlights ongoing legal questions about whether AI companies can use copyrighted material for training without permission or compensation to rights holders.

Beyond copyright issues, the practice raises preservation concerns. Rare book experts worry that unique historical artifacts—some centuries old, with handwritten annotations and physical characteristics that convey information beyond the text itself—are being permanently lost once converted to digital files and destroyed.

Why it matters

The tension between AI development and cultural preservation represents a new front in debates over AI training practices. While companies argue digitization makes texts more accessible, the permanent destruction of physical artifacts eliminates historical context, provenance, and physical characteristics that scholars value. The practice also suggests AI companies may be willing to sacrifice irreplaceable cultural heritage for marginal efficiency gains in data acquisition.

Industry Response

Vendors and rare book specialists are increasingly cautious about sales that might result in destruction rather than preservation. The disclosure has intensified scrutiny of how AI companies source training data and whether current practices adequately balance technological advancement with cultural stewardship responsibilities.

These details were first reported by The Guardian based on unsealed court documents in the ongoing Anthropic lawsuit.