Judge says Meta can use copyrighted work to train AI model

U.S. District Court Judge Vince Chhabria on Wednesday ruled in favor of the tech company after a group of authors claimed it had illegally used their writing to train its Artificial Intelligence program called Llama. Meta adhered to fair use doctrine under copyright law and could use the material, Chhabria ruled in a summary judgment. However, the judge wrote that his decision should not be taken as a sweeping ruling that tech companies could use copyrighted material to train their large language models. Instead, he said he made the judgment because the authors’ group focused on the wrong arguments and did not prioritize their suggestion that Meta’s actions could create a generative AI product that could flood and dilute the literary market with similar works.

What were some of the authors’ claims? The plaintiffs included well-known authors Sarah Silverman, Richard Kadrey, and Christopher Golden. They argued that Meta’s AI violated fair use because the model could reproduce excerpts of their books. The group also said Meta’s use of their work without permission diminished their ability to license their content for training. Chhabria said those arguments were not strong enough to support the case. While they said Meta did not have the right to use their content without permission, the tech company argued they adhered to fair use.

Have there been any similar cases recently? U.S. District Court Judge William Alsup on Monday ruled in favor of the company Anthropic’s use of copyrighted material to train its AI model. While he said it was legal for the company to use published books without an author’s permission, Alsup said a court would still hold a trial over claims that the company pirated books. Authors have brought similar lawsuits against OpenAI, Google, and other tech companies.

Dig deeper: Read Elizabeth Russell’s report about Disney and Universal Studios suing Midjourney for copyright infringement.