Meta Platforms Inc. has secured a significant legal victory in a copyright lawsuit filed by a group of authors who alleged that the tech giant unlawfully used their books to train its generative AI model, Llama. On Wednesday, Judge Vince Chhabria of the U.S. District Court in San Francisco dismissed the authors' claims, stating that they had not provided sufficient evidence that Meta's AI system infringed their copyrights under existing U.S. law.
The lawsuit, initiated in 2023, accused Meta of utilizing copyrighted works, including pirated versions of books, without permission or compensation, as training data for Llama. The plaintiffs, a group of 13 authors including Sarah Silverman, Ta-Nehisi Coates, and Junot Díaz, argued that Meta's actions constituted "massive copyright infringement". They were represented by a legal team led by attorney David Boies. The authors sought a court declaration that Meta's use of their copyrighted material was unlawful.
Meta defended its practices by invoking the "fair use" doctrine, a key principle of U.S. copyright law that allows for certain transformative uses of copyrighted materials for purposes such as news reporting, research, and parody. Meta's legal team, led by attorney Kannon Shanmugam, argued that copyright laws protect the expression of ideas, not the ideas themselves. The company asserted that its AI training process is transformative and does not replicate or supplant the original works but instead creates something new. A Meta spokesperson stated that "fair use of copyright material is a vital legal framework for building this transformative technology".
Judge Chhabria's ruling emphasized that the authors had not presented a strong enough case to proceed, particularly regarding how Meta's AI could harm the market for their work. He noted that the legal boundaries of fair use in the AI context remain unsettled and will require better-developed cases to test them. While the judge acknowledged the potential harm to the creative sector from generative AI, he stated that the authors failed to provide sufficient evidence that the market for their works had been diluted. He suggested that creators reframe their legal arguments to focus on market dilution and unfair competition rather than direct copyright infringement.
The court's decision aligns with a similar ruling in favor of Anthropic, another AI company facing a copyright lawsuit. In that case, Judge William Alsup ruled that Anthropic's training of its Claude LLM on copyrighted works was also "fair use," emphasizing the "transformative" nature of AI training. However, Anthropic faces a separate trial for allegedly downloading the works it used to train Claude from pirated "shadow libraries".
The authors' legal team expressed disappointment with the ruling, with a spokesperson from Boies Schiller Flexner LLP stating that "despite the undisputed record of Meta's historically unprecedented pirating of copyrighted works, the court ruled in Meta's favor". They respectfully disagreed with that conclusion but declined to comment on whether they would appeal the decision.
This case is one of several copyright lawsuits brought by writers, news outlets, and copyright holders against major tech companies, including OpenAI, Microsoft, and Anthropic. These lawsuits challenge the widespread practice of using copyrighted material for AI training and whether it constitutes "fair use" under U.S. law. The rulings in the Meta and Anthropic cases could have significant implications for the future of AI development and the rights of copyright holders.
It is important to note that Judge Chhabria clarified that his ruling does not mean Meta's actions were lawful, but rather that the plaintiffs failed to make strong arguments. He cautioned that using copyrighted books to build tools capable of flooding the market with competing works may not always be protected by fair use, especially when such tools generate vast profits.
