Authors take legal action against Meta for AI’s direct text replication.

Meta is currently facing a class action lawsuit initiated by five prominent book publishers and an author, who claim the company has committed one of the largest infringements of copyrighted materials in history while training its Llama AI models. The plaintiffs—Macmillan, McGraw Hill, Elsevier, Hachette, Cengage, and author Scott Turow—assert that Meta has repeatedly copied their books and journal articles without obtaining permission.

The lawsuit alleges that Meta knowingly sourced copyrighted content from well-known piracy sites like LibGen, Anna’s Archive, Sci-Hub, and others, incorporating this material into its AI model. It also contends that Llama was trained using the Common Crawl dataset, which is reportedly filled with unauthorized copies of copyrighted works. Consequently, Llama is said to produce outputs that are verbatim or nearly identical to the original copyrighted material. For instance, when given a prompt from Cengage’s popular textbook, *Calculus: Early Transcendentals, 9th edition*, Llama reportedly reproduces the text word-for-word.

Several authors have previously taken legal action against Meta for copyright infringement, revealing internal discussions about how to address media reports suggesting the use of a dataset known to contain pirated content. In a related case last year, a federal judge ruled in favor of Meta, but clarified that this decision does not imply that the use of copyrighted materials for training language models is lawful.

Additionally, a group of authors has filed a lawsuit against Anthropic for copyright infringement. Although a federal judge determined that training AI models on legally purchased books without permission could be considered fair use, he permitted the authors to proceed with a class action lawsuit regarding the alleged piracy of millions of works by Anthropic. The company agreed to a $1.5 billion settlement last year to resolve the class action.

Turow and the publishers are seeking damages from Meta and are requesting that the court mandate the company to cease its allegedly illegal activities. They also want Meta to disclose a list of the books, journal articles, and other copyrighted works used to train its Llama AI models.

In response, a Meta spokesperson stated, “AI is powering transformative innovations, productivity, and creativity for individuals and companies, and courts have rightly found that training AI on copyrighted material can qualify as fair use. We will fight this lawsuit aggressively.”