A former Meta researcher has testified that he downloaded an AI training dataset via BitTorrent, including the pirated Books3 collection. According to various rightsholders, this happened while he worked at the company, and he later deleted his copy “[d]ue to all this litigation.” Meta, which admitted training AI models on Books3 but said it didn’t believe it torrented the dataset, disputes the researcher’s account.
Source: Ex-Meta Researcher Torrented Pirated Books Dataset