Meta LLaMA AI Tool Faces Copyright Lawsuit by Authors

A lawsuit has been filed against Meta Platforms Inc., the parent company of Facebook, accusing it of utilizing the work of well-known authors to train its latest artificial intelligence tool, Llama.

Authors Richard Kadrey, Sarah Silverman, and Christopher Golden allege that the model was trained by “copying massive amounts of text from various sources and feeding these copies into the model.”

In a recent court filing, the three prominent authors provided several reasons for alleging that Meta’s AI tool LLaMa copied their books. LLaMa is a large language model that takes in a series of words and predicts the next word to create text step by step.

“Plaintiffs allege “on information and belief” that Meta has “benefited financially” from this noncommercial release of LLaMA.”

The filing argues that Copyright law doesn’t cover facts or the way information is structured and presented in books like those of the Plaintiffs during training.

It also points out that the content they refer to constitutes a very small fraction of the entire training dataset. “Even accepting Plaintiffs’ allegations, their books comprised less than a millionth of the training data,” it stated.

It highlights that using texts to train LLaMA for language modeling and creating original content resembles Google’s use of books for its search tool. In the case of Authors Guild v. Google, Inc., the practice was determined to be fair use.

Meta states that reviewing a research paper is the way for those interested in understanding how LLaMa was trained.

It also states that LLaMA will be accessible to academic researchers, individuals linked with government, civil society, academia, and industry research labs globally. However, primarily for studying, researching, and developing purposes.