Meta accused of using 82TB pirated books for AI training

Meta is facing renewed accusations of allegedly using 82 terabytes of pirated digital books to train its artificial intelligence models without authorization. Critics, citing a Tom's Hardware report, claim the company utilized copyrighted literature obtained through illicit channels rather than licensed datasets. This allegation intensifies ongoing scrutiny regarding how major technology firms acquire training data for large language models. Intellectual property advocates argue that using unauthorized copies constitutes copyright infringement regardless of transformative use claims. Meta has not publicly addressed these specific allegations regarding the 82-terabyte dataset. The controversy highlights the persistent tension between AI development speed and intellectual property rights enforcement. Legal experts suggest such claims could influence pending litigation and future regulatory frameworks governing generative AI training data. The outcome may determine whether AI companies must retroactively audit or purge unlicensed content from their foundational models.