On September 30, 2026, the U.S. Court of Appeals for the Third Circuit issued a landmark ruling in Thomson Reuters v. ROSS Intelligence, affirming a district court's grant of partial summary judgment in favor of Thomson Reuters and holding that ROSS Intelligence infringed copyrighted Westlaw headnotes when it used them to train its AI-powered legal research engine. The appellate decision, reported by IPWatchdog, LawNext, and multiple legal outlets, marks the first time a federal circuit court has addressed the fair use defense in the context of artificial intelligence training, and its reasoning will shape how courts evaluate similar claims in the wave of AI copyright litigation that is now moving through the federal judiciary. For personal injury law firms, the ruling is significant because it addresses the foundational question of whether AI developers can freely use copyrighted legal content to build competing tools, and because the court's analysis of market substitution and transformative use provides a framework that will influence not only copyright litigation but also the competitive landscape for legal AI products that PI firms rely upon for research, analysis, and case preparation.
The factual background of the case involves ROSS Intelligence's use of more than 2,000 copyrighted headnote summaries from Thomson Reuters' Westlaw platform to train its non-generative AI legal search engine. Headnotes are the concise summaries of legal principles that appear at the beginning of case reports in Westlaw, organized according to Thomson Reuters' proprietary Key Number System, a hierarchical taxonomy that classifies legal issues by topic and subtopic. ROSS, a legal research startup that ceased operations in 2021 due in part to the costs of this litigation, argued that its use of the headnotes constituted fair use under Section 107 of the Copyright Act, claiming that the training use was transformative because it served a different purpose than Thomson Reuters' original publication. The Third Circuit rejected this argument, affirming the revised summary judgment opinion of U.S. Circuit Judge Stephanos Bibas, who presided over the case in the District of Delaware while sitting by designation.
The appellate court's legal analysis focused on two issues that will be central to future AI copyright cases. First, the court affirmed that both the Westlaw headnotes and the Key Number System met the threshold of originality required for copyright protection, rejecting ROSS's challenge to the validity of Thomson Reuters' copyright registrations. This holding establishes that structured legal content, including taxonomies, summaries, and organizational systems, enjoys copyright protection independent of the underlying judicial opinions, which are in the public domain. For PI firms that use AI research tools, this distinction is critical because it means that the value-added content that legal publishers create, headnotes, annotations, cross-references, and classification systems, is protectable intellectual property, and AI developers that scrape or copy this content without authorization are exposed to infringement liability regardless of whether the underlying cases themselves are public domain.
Second, and more consequentially, the court rejected ROSS's fair use defense by applying the four-factor test established by the Supreme Court's 2023 decision in Andy Warhol Foundation for the Visual Arts v. Goldsmith. The Third Circuit concluded that ROSS's platform was a market substitute for Westlaw, failing the fourth factor of the fair use analysis, which examines the effect of the use on the potential market for the copyrighted work. The court determined that by using Thomson Reuters' proprietary headnotes to build a competing legal research tool, ROSS appropriated the value of the material and usurped Thomson Reuters' opportunity to license its data or enter the AI-driven market itself. The court also found that the use was not transformative under the Warhol framework because ROSS's purpose, building a legal research tool, was substantially similar to Thomson Reuters' purpose in creating the headnotes. The Third Circuit included a footnote distinguishing this case from ongoing generative AI litigation, noting that ROSS's system was a non-generative search engine and that the copying occurred before the widespread emergence of modern large language models, but the core reasoning about market substitution and transformative use will be directly applicable to generative AI cases.
For personal injury law firm leadership, the Third Circuit's ruling carries three practical implications. First, the decision establishes that legal publishers' copyrighted content, including headnotes, summaries, and organizational systems, is protectable against unauthorized use in AI training, and PI firms that evaluate legal AI vendors should inquire about the provenance of the training data, because tools built on scraped or unauthorized content may face shutdown orders, licensing disputes, or degraded functionality if the underlying copyright claims are enforced. Second, the court's emphasis on market substitution as the dispositive fair use factor suggests that AI tools that compete directly with the source of their training data are unlikely to prevail on fair use defenses, and PI firms should be cautious about adopting AI research tools from startups that have not established clear licensing relationships with the publishers whose content they use, because these tools may be legally precarious and could disappear if infringement litigation succeeds. Third, the appellate-level precedent, the first of its kind in AI copyright litigation, provides a template that other circuits will consider in pending cases involving generative AI models trained on copyrighted books, images, and music, and PI firms should monitor how this reasoning is applied to the legal AI tools they use, because a broad interpretation of the market substitution principle could limit the availability of low-cost or open-source legal research alternatives while strengthening the market position of established publishers like Thomson Reuters and LexisNexis. As the Third Circuit delivers the first appellate ruling on AI and fair use, the decision is a reminder that the legal framework for AI development is being written in real time, and personal injury firms must choose their technology partners with an understanding of the copyright risks that can determine whether a tool survives or is shut down by litigation.
