Elsevier joins lawsuit accusing Meta of using copyrighted research to train Llama

Elsevier has joined a class-action lawsuit against Meta that accuses the company of reproducing copyrighted works while developing its Llama large language model. The case, filed on 5 May in the Southern District of New York, also names Meta chief executive Mark Zuckerberg as a defendant, along with book publishers Hachette and Macmillan and US fiction author and lawyer Scott Turow.

Elsevier vs Meta: first science publisher sues over scraped research papers.

Zdroj: https://www.nature.com/articles/d41586-026-01481-0

Elsevier publishes thousands of journals, including Cell and The Lancet. The publishers allege that Meta obtained and reproduced copyrighted material in training Llama, including scientific abstracts and paywalled papers. The lawsuit says Meta used the Common Crawl data set, which the plaintiffs argue likely contained unauthorized copies of copyrighted works, and also downloaded or torrented material from sites including LibGen and Sci-Hub.

The case follows similar lawsuits brought by authors and media companies, including The New York Times, over the use of copyrighted material in AI training. Some of those cases have been settled, but no clear precedent has yet been established on whether training an LLM on copyrighted works is legal.

A Meta spokesperson said the company would “fight this lawsuit aggressively” and has suggested it will argue that training on copyrighted documents qualifies as fair use under US copyright law.


Zdroj: Nature News

Pôvodný článok: https://www.nature.com/articles/d41586-026-01481-0


Môže sa Vám ešte páčiť...

Pridaj komentár

Vaša e-mailová adresa nebude zverejnená. Vyžadované polia sú označené *