Appeals court: training an AI on Westlaw notes is not fair use

A US appeals court has rejected the argument that training an AI system on a publisher's copyrighted material is fair use, upholding Thomson Reuters' win over the legal research startup Ross Intelligence.

The Philadelphia-based Third US Circuit Court of Appeals ruled on Tuesday, September 29, 2026, as Reuters reported. The court rejected Ross' argument that its AI-powered search engine made fair use of material from Thomson Reuters' Westlaw platform, in what Reuters describes as the first ruling of its kind by a US appeals court in the wave of copyright cases over AI training. Law360 puts it the same way: the first US appellate ruling to address whether AI training can be fair use. The court's reasoning is sealed, according to Reuters and Legal IT Insider.

Thomson Reuters sued in 2020. It alleged that Ross copied Westlaw's headnotes, the short summaries of points of law attached to court opinions, and used them to train a competing legal search tool, and that thousands of headnotes were involved. Ross shut its platform down in 2021, citing the cost of the litigation. A federal judge in Delaware rejected Ross' fair use defence, and the line Reuters carries from that ruling is the shape of the whole case: "Ross took the headnotes to make it easier to develop a competing legal research tool," the judge said. "So Ross's use is not transformative."

A Thomson Reuters spokesperson said the company was pleased with the ruling and "firmly believes that respecting copyright is essential for fostering innovation while protecting the intellectual property that fuels fiduciary-grade AI solutions." Spokespeople and an attorney for Ross did not immediately respond to Reuters' requests for comment.

Reuters notes a distinction that matters for anyone reading the decision as a verdict on generative video. This case does not involve generative AI. Ross used the material to train a search engine that returned court opinions to a query, not a system that produces new text, images or footage, and the appeals court's reasons for its decision are sealed rather than published.

The stakes are familiar from the other direction. Reuters sets out the two positions the wider litigation runs on: tech companies argue that generative AI systems make fair use of the copyrighted material they are trained on by transforming it into something new, while copyright owners say the companies use their work to generate competing content that threatens their livelihoods. Authors, news outlets and music labels have filed dozens of suits on that question.

What this ruling settles is therefore narrower than the coverage of it. A US appeals court has now upheld one infringement finding against one AI training use, on facts about legal headnotes, with its reasoning under seal. For the video and audio model makers whose tools creators rely on, the open question is untouched: whether training a generative system on copyrighted footage, film or music is fair use has still not been answered in reasoning by any US appeals court.

The Third Circuit's reasoning is sealed, so how far the ruling reaches beyond Westlaw's headnotes is not yet on the public record.