Power up. From supercharging employee productivity and streamlining processes to inspiring innovation, Microsoft’s AI is designed to help you build the next big thing. No matter where you're starting, push what's possible and build your way with Azure's industry-leading AI. Check it out.
Can a group of high-profile authors write history when it comes to AI and copyright law?
Sarah Silverman, Ta-Nehisi Coates, and Junot Díaz are among those suing Meta in a copyright infringement case that could set an important precedent for how the law treats the training of generative AI systems.
Federal District Judge Vince Chhabria presided over a San Francisco courtroom hearing in that case last week, questioning lawyers on both sides about their respective arguments for hours.
The authors, represented by a legal team led by attorney David Boies, argue that Meta’s use of peer-to-peer file sharing sites to access “millions” of copyrighted works, including the authors’ books, constitutes “massive copyright infringement.”
Meta’s own motion argues that its training process constitutes fair use, a copyright law doctrine that allows for certain transformative uses of copyrighted materials for purposes like news reporting, research, and parody.
The case is one of several in the courts right now between GenAI companies and the media companies, publishers, and artists whose work was used to train models. Because these AI companies are similarly arguing that their training constitutes fair use, Chhabria’s decision could be a major signpost for how the law treats these issues.
Market “obliteration”: Much of Chhabria’s questioning of both sides focused on whether or not the market for books or other works is harmed by GenAI models, one of four key factors in deciding fair use. Chhabria said he would need to see proof of this market harm in order to rule in favor of the authors.
“I’ve been assuming that these models will be capable of producing competing works en masse, to the point that it will obliterate the market for certain copyrighted works,” Chhabria told Boies. “I think to rule in your favor, I need you to show that you’ve made that an issue in the case.”
Swift justice: Meta attorney Kannon Shanmugam argued that copyright laws don’t protect a creator’s style or ideas, but rather the expression. If a singer mimicked Taylor Swift’s style, for example, and became more popular than the pop star—“That’s impossible,” Chhabria jokingly interjected—that would not be protected by copyright law.
“But what about the next Taylor Swift?” Chhabria asked, arguing that mass generation of AI music would have “dramatically changed” the market for up-and-coming artists.
In his back-and-forth with Boies, Chhabria also questioned whether something like a memoir by Sarah Silverman would compete for customers with Llama outputs. Chhabria queried, too, the idea of harm to a secondary market for licensing AI training content. Without establishing that AI hurts book sales, “I think that is circular, as the other side says,” he said.
Boies and his team focused much of the motion on the piracy aspect. But the judge seemed less interested in that as a deciding point in the fair use question. “Seems kind of messed up,” Chhabria agreed. “But the question, as the courts tell us over and over again, is not whether something is messed up. The question is whether it’s copyright infringement, whether it’s fair use.”
The discussion touched on plenty of other aspects of copyright and AI during the three-hour hearing: the creative merit of AI prompting, the potential of AI to displace various types of copyrighted works, ideas versus expression.
“Very interesting. I’ll issue a ruling later today,” Chhabria said at the end of the hearing. “Just kidding! I will take a lot longer to think about it.”