Simply because Meta admitted to torrenting a dataset of pirated books for AI training functions, that does not essentially imply that Meta seeded the file after downloading it, the social media firm claimed in a court filing this week.

Proof as a substitute reveals that Meta “took precautions to not ‘seed’ any downloaded recordsdata,” Meta’s submitting mentioned. Seeding refers to sharing a torrented file after the obtain completes, and since there’s allegedly no proof of such “seeding,” Meta insisted that authors can’t show Meta shared the pirated books with anybody through the torrenting course of.

Whether or not or not Meta truly seeded the pirated books may make a distinction in a copyright lawsuit from guide authors together with Richard Kadrey, Sarah Silverman, and Ta-Nehisi Coates. Authors had beforehand alleged that Meta unlawfully copied and distributed their works by AI outputs—an more and more widespread criticism that up to now has barely been litigated. However Meta’s admission to torrenting seems so as to add a extra easy declare of illegal distribution of copyrighted works by unlawful torrenting, which has lengthy been thought-about established case-law.

Authors have alleged that “Meta intentionally engaged in one of many largest knowledge piracy campaigns in historical past to amass textual content knowledge for its LLM coaching datasets, torrenting and sharing dozens of terabytes of pirated knowledge that altogether comprise many tens of millions of copyrighted works.” Separate from their copyright infringement claims opposing Meta’s AI coaching on pirated copies of their books, authors alleged that Meta torrenting the dataset was “independently unlawful” below California’s Laptop Knowledge Entry and Fraud Act (CDAFA), which allegedly “prevents the unauthorized taking of knowledge, together with copyrighted works.”

Meta, nevertheless, is hoping to persuade the courtroom that torrenting will not be in and of itself unlawful, however is, moderately, a “widely-used protocol to obtain giant recordsdata.” Based on Meta, the choice to obtain the pirated books dataset from pirate libraries like LibGen and Z-Library was merely a transfer to entry “knowledge from a ‘well-known on-line repository’ that was publicly out there by way of torrents.”