AI firms shred millions of books after using them for training data

AI companies are buying up physical books through middlemen to train large language models, then shredding them, according to a new report. The practice involves secretly acquiring millions of books to build training datasets. This raises questions about copyright and the environmental impact of destroying physical media.