Last week the Court of Appeals for the Third Circuit, which oversees the districts of Delaware, New Jersey, Pennsylvania, and the Virgin Islands, upheld and further underlined a lower court ruling in the ongoing AI and copyright case of Thomson Reuters v. ROSS Intelligence Inc. At Courthouse News Service , Jackson Healy explains that the ruling found that materials which are changed with enough “creative spark” can be copyrightable, and therefore under existing copyright law cannot be fairly used by an AI to create a competing business.

This new decision mostly upholds a lower court’s ruling. There’s a good write-up of that original ruling by Yuanxiao Xu in the Authors Alliance that I’m drawing on heavily to understand the ruling and its impacts. From my reading, this seems to be a limited ruling that concerns a non-generative AI model, but the way the case treats copyright and a work’s transformation into an original could have major implications on how AI companies harvest data in the future.

Some background first. In 2020, Thompson Reuters sued an AI company, the now out-of-business ROSS Intelligence, over claims that one of their “AI-driven products to augment lawyers’ cognitive abilities” illegally ripped headnotes (descriptions at the top of judicial opinions containing summaries of legal issues) from Reuters’s legal research platform Westlaw. Crucially, ROSS was denied the ability to use Westlaw’s writing, and instead turned to a third party program to access and bundle the information it wanted.

ROSS then used this data to build a competing product, and Reuters sued over a breach of copyright. ROSS claimed fair use (a defense which has worked for other AI companies in the past), but in 2025, a judge ruled that ROSS using the headnotes was a breach of copyright law. The judge said that even though Westlaw is drawing initially on non-copyrightable judicial opinions for their headnote text, their arrangement and curation of the text had enough of a “creative spark” to be legally considered original work. To quote the judgment , “it still has the minimum ‘spark’ of originality.”

ROSS appealed, and this led to the most recent ruling in the Third Circuit by U.S. Circuit Judge Tamika Montgomery-Reeves, who affirmed the lower court’s findings: Westlaw’s headnotes possess “minimal ‘creative spark,’” and are therefore original. ROSS’s uses, on the other hand, were “minimally transformative at best” and in service of making a directly competing service. So, illegal.

Though “this case appears to concern the future of AI legal technology… In truth, this is no more than an ordinary copyright case,” wrote the Judge Montgomery-Reeves . This is important to me, because as much as these tech companies claim to be upending everything with AI and claiming extraordinarily special privileges as a result, there is still legal gravity constraining them.

But the ruling is a mixed bag, especially for books and publishing.

Why could this be troubling for other challenges to AI going forward? To quote Xu again : “First, it blurs the line between fact and expression… Second, it expands copyright enforcement to intermediate copies… Third, it conjures up a new market for AI training data.” All of this means that the transformation from fair and unfair use cases when it comes to text is debatable enough to be an open lane for a firm with enough time and money. Defining a way to transform work into something original that can be copyrighted and protected perhaps points a way for a new market to launder media so that AI can use it.