Meta Secretly Trained Its AI on a Notorious Piracy Database, Newly Unredacted Court Docs Reveal.

In order to feed the inexorable development of artificial intelligence, it is essential to have the right material to supply the models, so that they can improve day by day. Considering the constant advancement of this technology, the valuable content needed to ‘feed’ the various artificial intelligences is rapidly disappearing, forcing companies to seek out what's needed through less legitimate methods.

Over the past year, there's been much talk about the legal actions, and otherwise, taken by authors and writers whose works have been used as ‘training’ material for AI models, without prior permission from the owners. This seems to be the case with Meta, one of the leading companies in the industry, which recently lost a legal battle with a group of authors following a complaint of copyright infringement.

According to a series of documents made public by the court, the company apparently used Library Genesis (LibGen), a famous library of pirated books, to train its AI language models.