A US judge on Wednesday handed Meta a significant victory against authors who accused the tech giant of copyright infringement for using their works to train its Llama artificial intelligence (AI) models without permission. San Francisco District Court Judge Vince Chhabria ruled that Meta’s use of these works was “transformative” enough to qualify as “fair use” under copyright law, marking the second such triumph for AI firms this week.
“Transformative” Use vs. Market Harm
However, Judge Chhabria’s ruling came with an important caveat: the authors could have presented a winning argument that by training powerful generative AI with copyrighted works, tech firms are creating a tool that could allow a vast number of users to directly compete with them in the literary marketplace.
Chhabria stated in his ruling, “No matter how transformative (generative AI) training may be, it’s hard to imagine that it can be fair use to use copyrighted books to develop a tool to make billions or trillions of dollars while enabling the creation of a potentially endless stream of competing works that could significantly harm the market for those books.”
This highlights a core tension in the ongoing debate: while AI companies argue that training on large datasets is necessary for innovation and constitutes fair use due to the “transformative” nature of the process, content creators worry about the potential for their markets to be undermined by AI-generated content.
Meta’s Response and the Authors’ Arguments
A Meta spokesperson welcomed the decision, stating, “Open-source AI models are powering transformative innovations, productivity and creativity for individuals and companies, and fair use of copyright material is a vital legal framework for building this transformative technology.”
In the specific case before Judge Chhabria, authors had sued Meta for downloading pirated copies of their works, including Sarah Silverman’s “The Bedwetter” and Junot Diaz’s Pulitzer Prize-winning novel “The Brief Wondrous Life of Oscar Wao,” to train the Llama AI. Judge Chhabria clarified that his ruling does not establish that Meta’s use of copyrighted materials for training is broadly lawful, but rather that “these plaintiffs made the wrong arguments and failed to develop a record in support of the right one.”
Another AI Fair Use Ruling and Key Distinctions
This decision follows a similar ruling earlier in the week. On Monday, a different federal judge in San Francisco, District Court Judge William Alsup, sided with AI firm Anthropic regarding training its Claude AI models on copyrighted books without authors’ permission. Alsup ruled that Anthropic’s training using bought or pirated books was permissible under the “fair use” doctrine, describing the technology as “exceedingly transformative” and comparing AI training to how humans learn by reading.
However, it’s crucial to note a significant distinction: while Alsup found the training of Claude to be fair use, he rejected Anthropic’s bid for blanket protection for its practice of downloading millions of pirated books to build a permanent digital library. Alsup ruled that the creation of such a library was not justified by fair use protections. This suggests that while AI training might be permissible, the underlying methods of acquiring and retaining copyrighted data for building permanent libraries could still face legal challenges.
These rulings underscore the complex legal landscape surrounding AI development and copyright law, with ongoing litigation from various content creators across different industries.
