AI developers purchase and destroy millions of physical books for training models
Tech firms are buying and scanning millions of physical books to train AI models, often destroying the printed volumes during the process.
Artificial intelligence companies are increasingly acquiring millions of physical books to train their generative AI models. While tech firms previously relied on digital copies—sometimes facing legal challenges and allegations regarding pirated material—the industry focus has noticeably shifted toward physical printed media. Digitizing these hard copies typically involves unbinding or destroying the printed volumes so that individual pages can be rapidly scanned and converted into data for model training. This trend reflects growing efforts by AI developers to secure massive text datasets while navigating ongoing copyright disputes over digital content scraping.
Sources: Iltalehti.