By Adam Levine

A Federal judge in Delaware this week offered one of the first U.S. Court judgments in the debate around artificial intelligence and intellectual property.

In his ruling Tuesday night, U.S. judge Stephanos Bibas said that Thomson Reuters had seen its copyright violated by an AI start-up called Ross Intelligence. Ross had indirectly used its content to help train an AI product.

While the case began in 2020, before ChatGPT kicked off the current wave of generative AI, there have been a plethora of such cases filed since 2023. If this decision were to be used as precedent, it could spell trouble for the red-hot AI trade on Wall Street.

AI models need to be trained on millions of gigabytes of data, and that material has to come from somewhere. Model makers like OpenAI have said that they use "publicly available" data.

AI models often rely on data sets, like the frequently used Common Crawl, that contain copyrighted materials. There are narrow allowances around copyrighted material called fair use. Absent action from the U.S. Congress, courts will have to decide whether training AI models falls under these exceptions.

News Corp, the owner of Barron's publisher Dow Jones, has sued Perplexity AI, a generative AI start-up, for copyright infringement.

In Thompson Reuters versus Ross, the Bibas denied the claims of fair use "None of Ross's possible defenses holds water," he wrote in the ruling. "I reject them all." Bibas is a judge for the Third Circuit Court of Appeals temporarily serving as a District Court judge.

Steven Lieberman, a lawyer at Rothwell Figg which represents news organizations in two separate cases against OpenAI and Microsoft, thinks Tuesday's decision sends a strong message to AI companies. "If they want to use copyrighted content to train their models, they have to pay for it," he told Barron's. "This decision is a warning shot that they should heed that it's not acceptable."

But there's plenty of disagreement about whether the case will reverberate across the broader AI landscape.

Mark Lemley, a professor at Stanford Law School who represents defendant Stability AI in a copyright case, told Barron's that he thinks that the judge's "analysis of fair use is simply wrong."

Still it may have broader significance, Lemley said: "While the court is careful to distinguish this from the 30+ pending generative AI cases, I fear that this decision will set a bad precedent that could imperil AI more generally."

"The decision relies on a misreading of the prior case law," added Michael Carroll, professor at American University Washington College of Law, told Barron's.

Rebecca Tushnet of Harvard Law School also sees errors in the ruling but cautioned that "reasonable people could likely disagree on whether they should have changed the overall result."

As to whether this case might be used as precedent, law professors were skeptical. "While plaintiffs in the other cases will be cheered by Thompson Reuters' win, I think courts elsewhere will look to the facts before them, and not give too much weight to the Ross decision," Pamela Samuelson, a professor at UC Berkeley School of Law, told Barron's.

Investors also took a generally blasé approach to the ruling. While copyright law could one day undermine the AI trade, tech stocks and AI heavyweights like Microsoft and Meta Platforms were little moved on the federal court ruling.

Unable to fundraise with the lawsuit pending, Ross ceased operations in 2021, so its court loss is unlikely to see an appeal.

Write to Adam Levine at adam.levine@barrons.com

This content was created by Barron's, which is operated by Dow Jones & Co. Barron's is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

February 13, 2025 14:09 ET (19:09 GMT)

Copyright (c) 2025 Dow Jones & Company, Inc.