AI Companies Engage in Destructive Scanning of Antique Books for LLM Training Data

This practice raises significant ethical questions about the preservation of cultural artifacts and the future of intellectual property.

• AI companies began purchasing and destructively scanning antique books in early 2024, with major players like Anthropic and Amazon leading the charge.
• Court documents revealed plans for scanning up to 2 million books every six months, raising alarms about the destruction of rare texts.
• Public backlash has emerged from authors and historians, but no regulatory actions have been taken to halt these operations.

• Legal loopholes: The first-sale doctrine allows companies to destroy physical books after scanning, creating a gray area in copyright law.
• Quality concerns: AI developers are seeking high-quality, pre-2022 texts to avoid the pitfalls of training on AI-generated content, which can degrade model performance.
• Market demand: There is a growing demand for pre-AI-era texts, leading to increased bulk purchases of rare and out-of-print books by AI firms.

In early 2024, Anthropic launched Project Panama, a systematic initiative aimed at acquiring and destructively scanning millions of antique and rare books. This project was driven by the need for high-quality training data for large language models (LLMs), particularly texts authored by humans before the rise of AI-generated content. The legal landscape surrounding digital archives has become increasingly fraught, with AI developers facing liabilities for using pirated materials. As a result, the acquisition of physical books has become a preferred method, as these texts fall under the first-sale doctrine, allowing companies to legally destroy them after scanning.

The revelations about this practice came to light through court documents in the Bartz v. Anthropic lawsuit, unsealed in January 2026. These documents outlined ambitious plans for scanning between 500,000 and 2 million books every six months, indicating a large-scale operation. Booksellers reported a surge in bulk orders for rare titles, as AI companies sought to build their libraries of pre-AI texts. Investigative reports, including one published by 404 Media on August 17, 2026, confirmed that shipments of these books were routed to an Amazon warehouse in Las Vegas, specifically designed for spine removal, scanning, and disposal.

The implications of this practice extend beyond the immediate destruction of books. It raises ethical concerns about the preservation of cultural heritage and the potential loss of unique historical texts. Authors, archivists, and historians have voiced their outrage, labeling the practice as destructive to cultural memory. Despite the backlash, there has been no documented regulatory response from governments, leaving the market dynamics largely unchallenged.

As AI companies continue to prioritize the acquisition of physical texts, the demand for pre-2022 books is likely to sustain. This trend could lead to a scarcity of rare books in the market, impacting collectors, researchers, and anyone interested in accessing historical knowledge. The long-term consequences of this practice may reshape how we view intellectual property and the preservation of cultural artifacts in the digital age.

• Authors: Concerned about the loss of their works and the implications for copyright.
• Archivists and historians: Alarmed by the destruction of cultural heritage and unique texts.
• Booksellers: Experiencing increased demand for rare and out-of-print books, but facing ethical dilemmas.
• Researchers and students: Potentially losing access to important historical texts that are being destroyed.

• Regulatory developments: Monitor any potential government actions or regulations aimed at protecting cultural artifacts and intellectual property.
• Market shifts: Watch for changes in the availability and pricing of rare books as demand from AI companies continues to rise.
• Public sentiment: Keep an eye on the growing backlash from authors and cultural institutions, which could influence future practices in the industry.