Would You like a feature Interview?
All Interviews are 100% FREE of Charge
“The only time they’ll get into trouble is if they incorrectly summarize a story and make it defamatory when it wasn’t before. That’s something that actually puts them at legal risk, especially if they don’t credit the original source clearly enough and people can’t easily verify the source,” he says. “If Perplexity’s edits make a story defamatory, then under the body of case law that interprets it, Section 230 doesn’t cover it.”
In one case observed by WIRED, Perplexity’s chatbot falsely claimed WIRED had reported that a specific police officer in California had committed a crime, while prominently linking to the original source. (“We’re upfront that our answers are not 100% accurate and may even be hallucinated,” Srinivas said in response to questions for a story published earlier this week. “But core to our mission is to continually improve accuracy and the user experience.”)
“Formally, I think this is a series of claims that could survive a motion to dismiss based on a series of theories,” Grimmelmann says. “I’m not saying they’ll ultimately win, but if the allegations of potential plaintiffs like Forbes, WIRED, the police officers, etc. are supported by facts, they could be proven, and if other facts are unfavorable to Perplexity, they could lead to liability.”
Not all experts agree with Grimmelmann. Pam Samuelson, a professor of law and information studies at the University of California, Berkeley, wrote in an email that copyright infringement is “using someone else’s expression in a way that undermines the author’s ability to obtain appropriate compensation for the value of the unauthorized use. Using a single sentence verbatim is probably not copyright infringement.”
Vammati Viswanathan, a faculty fellow at the New England School of Law, is skeptical that the summary meets the substantial similarity standard typically required for an infringement suit to succeed, but she doesn’t think it’s the end of the story. “This summary cannot be beyond doubt,” she said in an email. “I would argue that the summary should be enough to meet the standard for a motion to dismiss, especially given all the indications that there was, in fact, copying.”
But overall, she argues, focusing on the narrow technical merits of such claims may not be the right way to think about things, since technology companies can adjust their operations to respect the letter of outdated copyright law while flagrantly violating its objectives. She thinks an entirely new legal framework may be needed to correct market distortions and advance the fundamental objectives of U.S. intellectual property law, one of which is to help people financially benefit from and encourage the creation of original works like journalism, which in theory would benefit society.
“In my opinion, there are strong arguments to support the intuition that generative AI presupposes massive piracy,” she writes. “The first presuppositional question is, where do we go from there? And in the long term, there’s a big question about how creators and the creative economy can survive. Ironically, AI is teaching us that creativity is more valuable and in demand than ever before. But even as we recognize this, we see the potential for it to undermine and ultimately eviscerate the ecosystem that allows creators to make a living from their work. That’s the conundrum we must solve – now, not eventually.”