By Emilio B. Nicolas

The U.S. Court of Appeals for the Third Circuit issued the first federal appellate decision directly addressing copyright fair use in the AI-training context. On September 29, 2026, in a long-running and closely watched legal battle between Thomson Reuters and the now-defunct startup ROSS Intelligence, the Third Circuit affirmed a district court’s decision that Ross infringed the copyrights in Thomson Reuters’s Westlaw headnotes by using them to train an AI-driven legal research tool. The case is Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., No. 25-2153 (3d Cir.). The decision marks a significant legal development for AI developers and copyright owners, while leaving broader questions about AI training and copyright for future cases.

A Little Background

As previously reported, Thomson Reuters, the owner of Westlaw, sued Ross in Delaware federal court alleging that Ross used more than 2,000 Westlaw headnotes—summaries of key points of law and case holdings—to train its own competing AI-driven legal research tool. On February 11, 2025, Judge Stephanos Bibas granted Thomson Reuters’s partial motion for summary judgment on its direct infringement claim and rejected Ross’s fair use defense. Importantly, Judge Bibas noted that Ross’s AI was not “generative AI (AI that writes new content itself). Rather, when a user enters a legal question, Ross spits back relevant judicial opinions that have already been written. . . .That process resembles how Westlaw uses headnotes and key numbers to return a list of cases with fitting headnotes.”^[1] Ross then appealed the decision to the Third Circuit.

The Appeal and Decision

In its appeal, Ross argued that the headnotes were ineligible for copyright protection and that its use of the headnotes to train its AI-driven legal research tool was a fair use. Following oral argument on June 11, 2026, a three-judge panel of the Third Circuit issued the September 29 ruling: a one-page decision that simply “AFFIRMED” Judge Bibas’s February 11 order.^[2] The panel also issued an accompanying opinion under seal, which was made public the very next day.^[3]

In its opinion, the Third Circuit framed the dispute as “no more than an ordinary copyright case.”^[4] It then held that the Westlaw headnotes at issue met the minimal “creative spark” required for copyright protection, and that Ross’s use of the headnotes did not qualify as “fair use”—an affirmative defense to copyright infringement—because three of the four statutory fair use factors weighed against such a finding.^[5]

Importantly, the Third Circuit found that Ross’s use was not sufficiently “transformative” for fair use purposes. Courts assess the transformative nature of an allegedly infringing work by asking “whether the new work merely supersedes the objects of the original creation . . . or instead adds something new, with a further purpose or different character.”^[6] The Third Circuit determined that Ross’s use of the Westlaw headnotes to train its AI-driven legal research tool was “minimally transformative, at best,” because Ross used the headnotes for the same ultimate purpose as Thomson Reuters’s use: “to create and optimize a legal-research platform that helps users find responsive legal material.”^[7]

In a footnote, the Third Circuit distinguished this case from other high-profile AI-training cases such as In re: OpenAI, Bartz v. Anthropic PBC, and Kadrey v. Meta Platforms.^[8] Consistent with Judge Bibas’s earlier finding, the court noted that Ross’s AI platform was not generative AI, “meaning it would not create any new expression; it would only return text passages from preexisting judicial opinions.”^[9] The other cases, in contrast, involve generative AI, which, in the Third Circuit’s view, may function to generate original expression.^[10] The decision thus resolves a dispute over a non-generative legal research tool and Westlaw headnotes, and leaves generative AI for another day. Still, the “ordinary copyright case” label undersells the decision: as circuit-level precedent, it draws early lines about fair use that parties on both sides of pending and future AI-training disputes will be quick to cite.

Practical Implications for Using Third-Party Content to Train AI

Building on the earlier analysis, companies using third-party content to train AI should consider the content’s source, evaluate rights and licensing options, and assess whether the use or resulting product competes with or substitutes for the content owner’s offerings. The Third Circuit’s decision does not resolve how those considerations apply to generative AI or other materially different systems, but it underscores the value of addressing such issues before deployment.

Because fair use is a fact-specific inquiry, the decision also reinforces the need to analyze the particular technology, source material, justification for use, and market effects at issue in each case rather than treating the ruling as either a categorical bar on AI training or a general safe harbor. AI developers, AI users, and copyright owners should account for those factual distinctions as federal courts across the country work through an ever-growing number of AI copyright cases.

For now, the Third Circuit’s decision marks a milestone, not the finish line. The court’s fair use analysis—along with the district court rulings in Bartz and Kadrey and future decisions involving AI, generative or otherwise—will help shape the next phase of AI copyright law. Stay tuned for updates.

[1]  Thomson Reuters Enter. Ctr. GmbH v. ROSS Intel. Inc., 765 F. Supp. 3d 382, 398 (D. Del. 2025).

[2]  Thomson Reuters Enter. Ctr. GmbH v. ROSS Intel. Inc., No. 25-2153 (3d Cir. Sept. 29, 2026), ECF No. 211.

[3]  Id. at ECF No. 214 (“Slip Op.”). The opinion was originally filed under seal with directions for the parties to propose redactions by October 9, 2026. Id. at ECF Nos. 208, 210. The opinion was unsealed on September 30, 2026.

[4]  Slip Op. at 2.

[5]  See Slip Op. at 11, 27-28.

[6]  See Slip Op. at 15, 16-17 (quoting Andy Warhol Found. for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 528 (2023)).

[7]  Id. at 16-17.

[8]  See In re: OpenAI, Inc. Copyright Infringement Litig., No. 1:25-md-3143 (S.D.N.Y.); Bartz v. Anthropic PBC, 787 F. Supp. 3d 1007 (N.D. Cal. 2025); Kadrey v. Meta Platforms, Inc., 788 F. Supp. 3d 1026 (N.D. Cal. 2025).

[9]  Slip Op. at 5, 17 n.7.

[10] See id.

The opinions expressed are those of the author(s) and do not necessarily reflect the views of the firm, its clients, or any of its or their respective affiliates. This article is for informational purposes only and does not constitute legal advice. For questions related to AI copyright issues, please contact a member of the Intellectual Property practice.

Meet Emilio

Emilio B. Nicolas is a copyright, trademark, and media rights attorney in Jackson Walker’s Austin office. His practice involves litigation and transactional work, including IP and media rights protection, enforcement, and licensing for companies in the entertainment, media, and technology industries. Emilio has been recognized by The Best Lawyers in America as Lawyer of the Year – Austin, Copyright Law (2024, 2027), by Lawdragon 500 as one of the “Leading Litigators in America” (2023-2027), and by Austin Monthly as one of “Austin’s Top Attorneys” for Intellectual Property Rights (2021-2026).