With generative AI increasingly becoming a fact of daily life, you could sometimes forget that the technology’s legality—at least as it’s currently trained—is still very much an open question.
A host of copyright cases are wending their way through the justice system that will decide the future of intellectual property rights in the AI era, shadowing the constant hype with a steady trickle of legal decisions. But lawyers we’ve spoken with expect it to take years for many of these questions to be fully decided—likely by the Supreme Court. By that time, generative AI’s capabilities could also play a role in the decision.
We spoke with IP lawyers about a couple of recent developments in this legal battle and where it’s likely headed.
First big court date: Thomson Reuters won the first major AI copyright case in the US in February against a legal startup called Ross Intelligence. Thomson Reuters alleged that Ross infringed on its copyright by training its AI-powered legal search engine on headnotes from Reuters’ Westlaw legal research service.
The judge agreed with Reuters and dismissed Ross’s argument that the training was covered by the fair use doctrine, a legal principle that allows use of copyrighted material without permission in certain contexts, like news reporting, research, or parody.
Because many AI companies are hinging their copyright arguments on fair use, the decision was seen as potentially significant. But before tea leaf readers get too excited, there are some factors that may narrow the impact of this particular case, lawyers we interviewed said.
For one, the judge noted that this decision, which stemmed from a 2020 lawsuit, does not involve generative AI. Aaron G. Rubin, a partner at Gunderson Dettmer, told us that limits how much transformation happens between training data and output, a key factor in the fair use doctrine. Ross was also building a direct competitor to Reuters, Rubin said, causing potential harm to the copyright holder’s market—another negator of fair use.
“Juxtapose that with what the foundational models are doing: They are taking all of the internet—ungodly sums of information—and they are putting it into models so that anyone can prompt those models for output,” Rubin said. “The ultimate question of whether generative AI using copyrighted data to train models is a fair use or not, that’s a separate question.”
Sarah Bro, a partner at McDermott Will & Emery, echoed this interpretation. “It’s very narrow and maybe not as applicable to most of the systems that are being trained right now that are in the generative mode.”
The case did, however, demonstrate how judges might consider different factors of fair use when it comes to AI, like whether the ability of the copyright holder to license data to AI companies constitutes its own market, Bro said. If future judges were to grant that market, it could have major implications across all AI and fair use cases, though Bro said that was just one possibility.
“I’m just really interested to see how the courts continue to interpret these factors in this area, and how much the ability for a copyright owner to license its data for training purposes—how that’s really going to affect that market-impact factor, and if that’s going to really gobble up all of the fair use cases,” Bro said. “I don’t know that we’re going to get to that sort of extreme black and white holding, but it’s possible.”
Keep up with the innovative tech transforming business
Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech.
All the news that’s fit to train: The New York Times and other plaintiffs also won a victory in their case against OpenAI last month when a US District Judge allowed the bulk of their arguments to proceed, while also narrowing the case somewhat.
It’s one of the most prominent in a slew of major cases between media companies and generative AI labs—Dow Jones v. Perplexity, Advance Local Media v. Cohere, and Getty Images v. Stability are others. Yet in all of these cases, Rubin said judges have yet to weigh in on the question of fair use in training, instead opting for more research.
“The courts in all of those cases [said the fair use] question wasn’t, as they say, ripe yet,” Rubin said. “There needs to be discovery, there needs to be more fact-finding in these cases before the courts are going to rule on that. So we largely don’t know yet what any of these courts are going to say.”
Once they do weigh in, district court decisions may conflict, Rubin and Bro agreed, meaning appellate courts and ultimately, the Supreme Court will have to decide.
Vincent Allen, a partner at Carstens, Allen & Gourley, said that process will probably take years.
“I do think it’s going to be a jury determination, and ultimately, probably will go all the way to the Supreme Court, unless there’s some legislation by Congress that happens in the meantime…I don’t think we’re going to see that kind of legislation,” Allen told Tech Brew. “I think the courts are going to make a decision about whether this is fair use or not, but we’re many years out from that right now.”
Audience of nine: By the time the Supreme Court does eventually weigh in, generative AI technology might look completely different from it does today. And how widespread it actually becomes could make a difference in the calculus of that final decision, Kate Downing, a lawyer focused on AI licensing and the tech industry, told us.
“The history of copyright law and the history of disruptive technology is that almost every disruptive technology we’ve ever had, whether you’re talking about the VCR or the photocopier or the phonograph—every single one of these things, when they came out, we had no idea whether or not it violated copyright laws. There were pretty good arguments to say that they did or could,” Downing said. “And in many cases, the question really turned on, by the time this case gets to the Supreme Court, is this a technology that the Supreme Court justices are going to appreciate? Like, is it going to be something that they understand, that they benefit from?”
For AI companies, that could mean creating a product mainstream enough to cater to one key customer demographic: Supreme Court justices.
“One of the things I said to the AI researchers last year is, ‘Look, to the extent that you guys believe in this technology, that you want to see it flourish, that you want to have an open field in which to work, you really need to hurry up and invent something that the Supreme Court justices are going to appreciate,’” Downing said. “And I don’t think it’s going to be a crappy support chatbot. I don’t think they’re going to appreciate that. I think it’s got to be an AI that really does something for their quality of life.”