The Third US Circuit Court of Appeals has ruled that ROSS Intelligence's use of Thomson Reuters' copyrighted legal headnotes to train its AI research platform was not fair use, handing the publishing giant one of the most consequential copyright victories yet in the emerging law of artificial intelligence. The decision is the first from a federal appeals court to address fair use in the context of AI training — and the now-shuttered ROSS says it will ask the Supreme Court to review the case.
According to LawNext, which has followed the litigation closely, the Third Circuit panel entered judgment affirming Judge Stephanos Bibas's February 2025 summary judgment in favor of Thomson Reuters. The appeals court sided with the publisher on the two questions at the heart of the appeal: whether Westlaw's headnotes — the short editorial summaries that sit atop case opinions — are copyrightable in the first place, and whether ROSS's internal use of that material as training data for an AI legal search engine was protected by fair use. For more context on this story, see our ongoing artificial intelligence updates.
The answer to both questions was no. The opinion was initially issued under seal while the parties requested redactions, and it became public this week, as reported by Courthouse News and analyzed in depth by Reuters, which described the unsealed opinion as showing why Thomson Reuters won its landmark AI fair-use ruling.
Why the Court Sided With Thomson Reuters
ROSS Intelligence was a legal research startup that set out to build a search engine that could compete with Westlaw, Thomson Reuters' flagship legal research platform. To train and calibrate its competing product, the company obtained and used Westlaw headnotes — thousands of editorially written summaries of legal principles drawn from court opinions.
Judge Bibas, a former federal appeals judge sitting by designation on the district court, had ruled in February 2025 that the headnotes were sufficiently original to qualify for copyright protection, and that ROSS's use of them was not fair use. A central factor in his analysis was purpose: ROSS had not copied the headnotes to build a generative AI system, but to develop a legal research product aimed squarely at the same market Thomson Reuters serves. That competitive purpose weighed heavily against a finding of fair use.
By affirming on both issues, the Third Circuit left those rulings in place and closed off ROSS's primary defense. Thomson Reuters welcomed the outcome, with a company spokesperson telling Reuters that the company "firmly believes that respecting copyright is essential for fostering innovation while protecting the intellectual property that fuels fiduciary-grade AI solutions."
The First Appellate Fair-Use Ruling on AI Training
The decision marks the first time a federal court of appeals has addressed fair use in the context of AI training, a distinction that guarantees it will be cited — and contested — in every major AI copyright fight now working through the courts.
But the ruling comes with important caveats. As LawNext notes, the unique facts of the case may minimize its broader implications. ROSS copied Westlaw's material before the explosion of generative AI, and it did so to build a non-generative search product intended to compete directly with the plaintiff. That distinguishes the case from disputes where developers trained large language models on publicly available material for general-purpose systems.
Publishers and authors, who have spent years arguing that AI training on their works requires permission, treated the ruling as vindication. Publishers Weekly reported that the Association of American Publishers and the Authors Guild applauded the decision, signaling that rights holders intend to lean on the ruling in ongoing and future disputes.
ROSS Heads to the Supreme Court
ROSS, for its part, is not done. The company said this week that it will seek review from the US Supreme Court, according to LawNext. Yar Chaikovsky, a partner at White & Case representing ROSS, said in a statement that the company respectfully disagrees with the Third Circuit's decision.
"We respectfully disagree with the Third Circuit's decision and believe it creates continued uncertainty around the application of copyright law to AI model training," Chaikovsky said. "We intend to seek review by the Supreme Court to obtain much-needed clarity on these issues, which carry significant implications for innovation and the development of AI technologies."
The Supreme Court accepts only a small fraction of the petitions it receives, but the certiorari question here is unusually attractive: the country's leading technology companies are locked in litigation over the same legal question, and appellate courts have only begun to weigh in. A high-court grant would put the fair-use question for AI training on a path to a definitive national answer.
What It Means for the OpenAI, Anthropic and Meta Cases
Observers have suggested the case could ripple into the highest-profile copyright battles in technology — including suits involving OpenAI, Anthropic and Meta over the books and articles used to train their models. But legal analysts, including LawNext's coverage, caution that the analogy is imperfect.
ROSS built a narrow, non-generative product with copied material obtained for the purpose of competing with its source. Generative AI developers, by contrast, argue their systems produce transformative outputs and do not substitute for the works they train on. Courts on both coasts are still testing those arguments, and the Ninth Circuit and Second Circuit have yet to issue controlling rulings on generative training.
That leaves the practical effect of this week's decision narrow but real: developers that use copyrighted works in ways that replicate or compete with the original — particularly against the same customers — now have a federal appellate ruling against them. And with ROSS promising a Supreme Court petition, the case that produced the first appellate fair-use ruling on AI training may also produce the last word on it.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
