Thousands of books are being sought, bought and digitized by artificial intelligence companies, not to be read, but to serve as training material for language models. The process begins with the acquisition of large quantities of books in bulk.
After the purchase, the digitization process begins, transforming these books into data that can be used to train AI systems. This practice is known as the "metamorphosis" of books into training material.
The practice raises important questions related to copyright. Authors and publishers question whether the use of their works to train language models without authorization or compensation is legitimate.
AI companies defend that this type of use is covered by copyright exceptions, but legislation on this matter is still being debated and clarified in various countries.
