IP Law Daily, COPYRIGHT NEWS: Tech company seeks quick appeal on copyrightability, fair use questions in Westlaw headnotes dispute, (Apr 16, 2025)

By Thomas Long, J.D.

Ross Intelligence, which included purportedly copyrighted headnotes in training data for an AI-assisted legal research tool, asks Third Circuit to reverse summary judgment ruling that it infringed.

Ross Intelligence, Inc.—a tech startup that describes itself as a builder of “AI-driven products to augment lawyers' cognitive abilities”—is seeking quick resolution on appeal of key issues in a copyright infringement suit brought against it by the owners and operators of the legal research database Westlaw. In response to a determination by the federal district court in Delaware on summary judgment that Ross infringed copyrighted Westlaw headnotes by including them in training data for its AI-assisted search tool, and that Ross did not engage in protected fair use, Ross has asked the Third Circuit for permission to pursue an interlocutory appeal under 28 U.S.C. § 1292(b). Ross requests the appellate court to hold that (1) Westlaw headnotes lack the requisite originality for copyright protection and (2) the challenged use of the Westlaw content by Ross was “fair” under Section 107 of the Copyright Act, 17 U.S.C. § 107 (Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc., No. 25-8018 (3d Cir. Apr. 14, 2025)).

Copyright dispute. The dispute centers on the Westlaw legal research platform, owned and operated by Thomson Reuters Enterprise Centre GmbH, and Ross’s use of the database contents to create a “natural language search engine” using machine learning and artificial intelligence. Along with the content of non-copyrighted judicial opinions, statutes, and other government documents, Ross used headnotes authored by Thompson employees that accompany and summarize the key issues of court cases. Thomson asserts copyright ownership in its headnotes, and it contended that Ross’s conversion of Thomson’s copyrighted material into machine-readable data suitable for training its AI constituted copyright infringement. Thomson filed suit in the District of Delaware in May 2020.

Summary judgment decisions. In a September 2023 decision on both parties’ motions summary judgment, the district court decided that factual questions existed that required consideration by a jury, particularly with respect to whether the contents of the Westlaw database were copyrightable and whether fair use was a defense. were not issues that could be decided as a matter of law. In February of this year, the court unexpectedly revisited the summary judgment motions and significantly revised its conclusions. The court determined that Westlaw’s headnotes were original enough to be copyrightable, that Ross as a matter of law unlawfully copied some of the headnotes, and that Ross’s fair use and other defenses were unavailing.

Motion for interlocutory appeal. On March 18, Ross filed a motion for interlocutory appeal and for stay pending appeal. On April 4, in an order entered on the docket without written opinion, the district court granted the motion. In the order, the judge said, “Though I remain confident in my February 2025 summary judgment opinion, I recognize that there are substantial grounds for difference of opinion on controlling legal issues in this case. These issues have the potential to change the shape of the trial. I thus certify the following two questions to the Third Circuit: (1) whether the West headnotes and West Key Number System are original; and (2) whether Ross's use of the headnotes was fair use. A short opinion further explaining my reasoning will follow.”

On April 14, Ross filed a petition for certification under Section 1292(b) with the Third Circuit.

Controlling questions of law. Ross asserts that the case is appropriate for interlocutory appellate review because—as the district court recognized—the originality and fair use questions involve controlling issues of law on which there are substantial grounds for differences of opinion, and an interlocutory appeal will materially advance the litigation. In addition, resolving these questions would serve the national interest, Ross said. According to Ross, the case deserves the appellate court’s immediate attention because “the certified questions are of unique importance that raise issues of fundamental importance about artificial intelligence, copyright, and national security that are likely to arise time and again.”

Copyrightability of headnotes. Ross’s petition cites the well-known case of Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991) for the proposition that copyright validity is a matter of law, not fact. In Feist, the Supreme Court said that a work must be original to be copyright-protectable, and it explained that “Factual compilations … may possess the requisite originality” depending on the “selection” and “arrangement” of the facts. Because copyright protection extends only to components that are original to the author, such a copyright is “thin,” particularly when the author adds no new expression but “lets the facts speak for themselves.”

Ross notes that the district court’s two summary judgment orders contained different holdings regarding the originality of the Westlaw headnotes. In the first order, the district court said that a headnote that “merely copies” a judicial opinion is not copyrightable, and Ross “present[ed] evidence that Thomson Reuter’s protocols required headnotes to follow or closely mirror the language of judicial opinions.” The court therefore concluded that there was a genuine factual dispute as to originality. However, in the second opinion, the district court found that the headnotes constitute a compilation, and that “Thomson Reuters’s selection and arrangement of its headnotes easily” met the requirement for a “minimal degree of creativity” set forth by Feist. In addition, the court said that each headnote was an individually copyrightable work, even those that used language from an opinion “verbatim.” The district court reasoned that each headnote was “a carefully chosen fraction of the whole, identifying which words matter.” The selection of which judicial language was relevant for purposes of the headnote was “editorial expression,” in the court’s view.

According to Ross, both opinions are wrong. “Westlaw headnotes, which parrot judicial opinions as best they can, do not contain a modicum of creativity,” either individually or as a compilation, Ross argues. As a matter of law, the headnotes merely contain non-copyrightable facts. Ross cited Matthew Bender & Co. v. West Publishing Co., 158 F.3d 674, 680, 689 (2d Cir. 1998), in which the Second Circuit determined that alterations to case reports, even considered collectively, failed to “demonstrate sufficient originality and creativity to be copyrightable.” Ross contends that under the reasoning of that case, Thomson’s “routine” or “garden-variety” editorial selections do not contain the “creative spark” required for copyright protection. Ross asserts that “‘industry conventions’” and functionality dictate the Westlaw headnote editors’ decisions,” so the process of selecting text from the opinion lacks copyright-worthy creativity or originality.

Fair use. With respect to fair use, Ross again argues that the “two conflicting opinions” by the district court “reflect substantial disagreement on the application and balancing of the fair use factors.” First, Ross notes disagreements between the opinions about the transformative nature of Ross’s use of AI. In the first opinion, the court concluded that a jury must decide whether Ross engaged in “transformative intermediate copying.” In the second opinion, the district court stated that the caselaw regarding intermediate copying reached only copying of computer code, not prose. In those cases, intermediate copying of code was necessary for innovation, in the district court’s view, whereas copying headnotes was not necessary for Ross to build its AI search feature. The district court said in its second summary judgment order that Ross’s use was commercial, not transformative. Ross contends that both decisions were “doctrinally flawed,” because neither opinion adequately addresses the question of whether Ross’s AI search feature “adds something new, with a further purpose or different character” (citing the recent Supreme Court decision in Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508, 528 (2023)). Ross’s contention is that the district court failed to answer whether that “something new” existed when Ross “built a search tool that uses AI to match natural language questions with excerpts straight from judicial opinions.”

Ross also attacked the district court’s approach to the fourth statutory fair use factor: the effect on the market for the copyrighted work. Ross pointed to the relevance of the public benefits that derive from the use, and noted that the district court in its first opinion stated that the question of whether it was in the public benefit to allow AI to be trained with copyrighted material had to be answered by a jury. In the second opinion, the court changed its mind and said that the only thing that mattered was that Ross “meant to compete with Westlaw by developing a market substitute” and it failed to “put forward enough facts to show” that the “potential market for AI training data” does not “exist and would not be affected.” The court also said in that opinion that the “public has no right to Thomson Reuters’s parsing of the law.” According to Ross, the district court erred because there was no authority that actor four considered the potential effect on a hypothetical market. Ross called it “stunning” for the district court to require “innovators … to affirmatively show that they did not create a new-hypothetical market; and, in the alternative, prove that this hypothetical market did not affect an alleged competitor.” Ross also criticized the district court’s statements on the public interest, arguing that Ross was not providing Thomson’s “parsing” of the law, but rather “verbatim quotes from uncopyrightable judicial opinions devoid of any third-party content.” Ross further contended that increasing public access to the law with “good research tools” benefits society.

With respect to balancing, Ross took issue with the district court’s position that the first and fourth Section 107 factors are of the highest importance, and argues that “neither [district court] decision adequately engages a critical, and undisputed, fact: ROSS used.076% of Westlaw headnotes.” The district court failed to take into proper account this quantitative detail, according to Ross.

The case is No. 25-8018.

Attorneys: Mark S. Davies (White & Case LLP) for Ross Intelligence Inc.

Companies: Ross Intelligence Inc.; Thomson Reuters Enterprise Centre GmbH

MainStory: TopStory AINews Copyright TechnologyInternet DelawareNews NewJerseyNews PennsylvaniaNews