By Blake Brittain and Katie Paul
(Reuters) – A group of writers including Pulitzer Prize winner Michael Chabon sued Meta Platforms in San Francisco federal court on Tuesday, accusing the tech giant of misusing their works to train its Llama artificial-intelligence software.
Chabon, Tony-winning playwright David Henry Hwang and authors Matthew Klam, Rachel Louise Snyder and Ayelet Waldman in a lawsuit said Meta taught the Llama large-language model to respond to human text prompts with datasets that included pirated versions of their writings.
The same writers filed a similar proposed class-action lawsuit on Friday against ChatGPT maker OpenAI. The authors said in the OpenAI case that works like books and plays are particularly valuable for AI language training as the “best examples of high-quality, long form writing.”
A spokesperson for Meta declined to comment on the new lawsuit. An attorney for the writers did not immediately respond to a request for comment.
Meta and OpenAI were also sued for copyright infringement in July by a separate group of authors that includes comedian Sarah Silverman, part of a growing list of copyright cases against AI companies.
Meta published a list of datasets used to train its first version of the Llama model, which it released in February. The company did not disclose training data for its latest version, Llama 2.
Llama 2, the first large language model that Meta has made publicly available for commercial use, is free to use for companies with fewer than 700 million monthly active users.
The Llama 2 release was seen as a potential game-changer in the emerging market for generative AI software, threatening to upend the early dominance of players such as OpenAI and Google that charge significant amounts to use their models.
(Reporting by Blake Brittain in Washington and Katie Paul in New York; Editing by David Bario and Daniel Wallis)