Anthropic's Book Scanning Project Raises Concerns

Alleged Secret Scanning Initiative Surfaces

According to claims circulating on social media, AI company Anthropic allegedly conducted a covert project involving the physical scanning of millions of books. The tweet from International Cyber Digest alleges that the company spent billions of dollars on an initiative that required slicing the spines off books to facilitate high-speed scanning. The first image shows Anthropic's branding with a minimalist logo design, while the second reveals a pile of various books with visible titles including 'The Ten Thousand', 'Classic Speeches', 'The Future of Us', and others. If accurate, this would represent one of the most extensive book digitization efforts undertaken by a private AI company, though the destructive methodology raises significant questions about preservation ethics and access to rare materials.

Destruction of Rare and Heritage Books

The most controversial aspect of these allegations centers on the claimed destruction of rare and potentially irreplaceable books. The tweet explicitly states that 'even very rare ones' were subjected to this destructive scanning process. Traditional book scanning typically involves careful page-by-page photography that preserves the physical artifact. However, industrial-scale scanning operations sometimes employ guillotine cutters that remove book spines to feed pages through high-speed scanners. This process is irreversible and transforms bound volumes into loose pages. For common books with multiple copies in circulation, this may be acceptable, but applying such methods to rare editions, first printings, or heritage materials would constitute a significant loss to bibliographic history and physical book culture. The cultural implications extend beyond mere digitization to questions of stewardship.

Training Data for Large Language Models

The likely purpose behind such an extensive scanning operation would be to create training data for Anthropic's Claude AI models. Large language models require vast text corpora to develop language understanding and generation capabilities. Books represent high-quality, edited, long-form content that is valuable for training purposes. However, the methods used to acquire this data matter significantly from both ethical and legal perspectives. Many AI companies have faced criticism and lawsuits over training data sourcing, particularly regarding copyright and author compensation. If Anthropic indeed pursued destructive scanning of millions of volumes, it would represent an unprecedented physical commitment to data acquisition. The billions allegedly spent would reflect both the acquisition costs of the books themselves and the industrial infrastructure required for such large-scale processing operations.

Implications for Access and Competition

A particularly concerning claim in the tweet is that this process means 'no other AI, and no human, can access' these materials. If rare books were destroyed in the scanning process, their physical forms would indeed be lost permanently. This raises questions about whether Anthropic created exclusive digital archives that provide competitive advantages over other AI companies while simultaneously removing materials from public access. In the broader AI industry, access to unique training data has become a significant competitive moat. Companies that can train on distinctive, high-quality datasets may produce superior models. However, achieving this advantage through destruction of cultural materials would represent a troubling precedent where corporate AI interests supersede preservation of human heritage and equitable access to knowledge resources for researchers and the public.

Industry Response and Verification Needed

As of this report, these claims remain unverified allegations from a single social media source. Neither Anthropic nor independent book preservation organizations have confirmed the existence of such a program. The images show only Anthropic branding and a stack of various books, which do not independently prove destructive scanning occurred. However, the allegations are specific enough—mentioning billions in spending and millions of books—to warrant serious investigation. If true, this would represent a major scandal in the AI industry with implications for how companies source training data. The bibliographic and archival communities would likely demand accountability and policy changes. For now, stakeholders await official response from Anthropic and corroborating evidence from other sources before drawing definitive conclusions about this alleged secret project.