Source: The Verge
Meta's use of copyrighted books to train its AI models without permission or compensation has moved from industry complaint to legal liability, with publishers arguing the company copied text "word-for-word" into its training datasets. The lawsuit exposes a widening gap between what tech companies claim is "fair use" research and what copyright holders—who already lost control of their digital distribution to Amazon—see as theft of their core asset. If publishers win, Meta and other AI labs would need to negotiate licensing deals or exclude books from training sets entirely, raising the cost of large language model development.