Key facts
- Independent booksellers worldwide report an increase in bulk orders for secondhand books.
- Purchases are suspected to be linked to AI companies scanning texts for training large language models.
- A US court ruled that scanning texts for AI training is a transformative use.
- The court ruling determined that such scanning is not a copyright violation.
- Ethical concerns persist regarding the destructive scanning of books.
- The practice raises questions about book destruction and literary preservation.
Independent booksellers across the globe are observing an unprecedented increase in bulk orders for secondhand books. These unusual purchasing patterns have led to growing suspicions that the books are being acquired by artificial intelligence companies. The primary concern is that these companies are destructively scanning the texts to train their large language models (LLMs). This development comes in the wake of a significant US court ruling that classified the scanning of copyrighted material for AI training as a transformative use, thereby not constituting a copyright violation. Despite the legal precedent, ethical questions surrounding the practice persist. Critics and booksellers alike express unease about the potential for irreversible damage to literary works during the scanning process. The long-term implications for book preservation and the accessibility of physical texts are subjects of ongoing debate.