Table of Contents
Artificial intelligence models require a massive amount of data to function effectively, but the origin of this data is sometimes controversial. A recent court case raises questions about Meta’s use of copyrighted books to train its AI model, Llama. Added to this are accusations involving Guillaume Lample, co-founder of Mistral AI, who allegedly played a role in this case.
The 3 key facts not to miss
The writers involved in the “Kadrey v. Meta Platforms Inc.” case accused Meta of using pirated works for the development of its AI model, Llama. These works were allegedly downloaded from Library Genesis, an illegal online library hosting books and scientific articles. Mediapart revealed that Meta allegedly used these resources without authorization, sparking debates on the ethics of such practices.
The trial verdict, delivered in June, was however favorable to Meta. The judge argued that the plaintiffs failed to demonstrate that Meta’s use of the books caused them harm, based on the principle of “fair use” which allows, under certain conditions, the use of protected works.
Guillaume Lample, before co-founding Mistral AI, worked within Meta’s AI team. Mediapart claims, based on documents and emails, that Lample was a key player in the decision to use LibGen data for exploratory AI purposes. The plaintiffs claim he downloaded 70 terabytes of protected data.
This revelation has drawn attention to data collection practices in AI and raises questions about the legality and ethics of using these resources.
Although the court ruled in favor of Meta, the plaintiffs’ lawyers expressed disagreement with this decision, arguing that the scale of the piracy was unprecedented. Meta, for its part, maintained that “fair use” is an essential legal framework for the development of innovative technologies.
Mistral AI, the company co-founded by Lample, stated that it only uses public, licensed, or synthetically generated data. Neither Meta nor Lample wished to respond to Mediapart’s questions.
Meta Platforms Inc., formerly known as Facebook, is an American multinational technology company founded by Mark Zuckerberg. It is primarily known for its social media platforms but has diversified into the field of artificial intelligence with the development of models such as Llama.
Mistral AI, on the other hand, is a French artificial intelligence company co-founded by Guillaume Lample. It is valued at 11.7 billion euros and focuses on the development of advanced AI technologies while complying with data usage regulations.