Image source: Glenn Carstens-Peters/unsplash.com
In the lawsuit filed by author Richard Kadrey against Meta✴, book authors and content copyright holders argue that Meta✴ was a “willing participant in an illegal peer-to-peer network” and operated without prior licensing agreements. The company used the BitTorrent protocol, which involves simultaneously distributing content to other users, to download data from shadow libraries. According to the prosecution, this amounts to distributing pirated material.
Meta✴, for its part, said in a court filing this week that while it downloaded 82 TB of copyrighted content from dark web libraries to train its LLaMA AI models, its staff “took precautions not to ‘see’ the downloaded files.”
The company’s defense is that there is currently no evidence that it shared downloaded content during the torrenting process. However, Michael Clark, Meta✴’s project manager, told the court that the torrent’s configuration parameters were changed “in such a way as to minimize seeding.” When the judge asked why Meta✴ chose to minimize seeding, the company’s lawyers invoked the attorney-client privilege, allowing Clark to leave the question unanswered.
Additionally, Meta✴ has not yet responded to questions about whether it knew that data was being shared with other users while downloading content from shadow libraries.