Thomson Reuters v. ROSS Intelligence: AI & Copyright
Title
Thomson Reuters Enterprise Centre GmbH & West Publishing Corp. v. ROSS Intelligence Inc.: Copyright Protection and Fair Use in AI Training.
Case Details
Case Name: Thomson Reuters Enterprise Centre GmbH & West Publishing Corp. v. ROSS Intelligence Inc.
Court: United States District Court for the District of Delaware
Judge: Hon. Stephanos Bibas, sitting by designation
Date of Judgment: 11 February 2025
Citation: No. 1:20-cv-00613-SB, 765 F. Supp. 3d 382 (D. Del. 2025)
Case Number: No. 1:20-cv-00613-SB
Introduction
The rapid development of Artificial Intelligence has created significant legal questions concerning the use of copyrighted works as training data. The decision in Thomson Reuters Enterprise Centre GmbH & West Publishing Corp. v. ROSS Intelligence Inc. is an important United States decision concerning the use of copyrighted legal research material to develop an AI-powered legal research tool. Thomson Reuters alleged that ROSS Intelligence had used material from its Westlaw platform, particularly editorially created headnotes and the West Key Number System, without authorisation to train its competing legal research product. The case required the court to consider whether such copying constituted copyright infringement and whether it was protected by the doctrine of fair use. In February 2025, the District Court granted partial summary judgment in favour of Thomson Reuters and rejected ROSS’s fair-use defence. The decision is particularly significant because it addresses the relationship between copyright protection, commercial competition and AI training.
Facts of the Case
Thomson Reuters operates Westlaw, a major legal research platform containing judicial opinions and other legal materials. Westlaw also contains editorial content, including headnotes, which summarise particular legal principles found in judicial opinions. These headnotes are organised through the West Key Number System, an editorial classification system developed by West Publishing.
ROSS Intelligence was developing an AI-powered legal research tool intended to compete with Westlaw. According to Thomson Reuters, ROSS did not have permission to use Westlaw’s copyrighted editorial material for developing its product. After Thomson Reuters refused to license the relevant material, ROSS allegedly obtained copies of Westlaw headnotes through a third-party company known as LegalEase. LegalEase created “Bulk Memos” containing legal questions and corresponding answers based upon Westlaw headnotes. ROSS used this material to develop training data for its legal research system.
Thomson Reuters subsequently filed a copyright infringement action against ROSS in the United States District Court for the District of Delaware. Thomson Reuters argued that ROSS had copied protected elements of Westlaw in order to create a competing legal research product. ROSS denied liability and relied, among other arguments, upon the doctrine of fair use under United States copyright law. It argued that its use of the material for developing its AI system was sufficiently transformative and should therefore be protected.
The litigation ultimately focused on whether the copying of Westlaw’s editorial headnotes for the development of ROSS’s competing legal research tool constituted copyright infringement and whether the use qualified as fair use. The court’s February 2025 decision addressed these issues through partial summary judgment.
Issues before the court
The principal issues before the court were:
Whether Thomson Reuters’ Westlaw headnotes contained sufficient original expression to receive copyright protection?
Whether ROSS had copied protected portions of Thomson Reuters’ copyrighted material?
Whether ROSS’s use of the copyrighted material to develop its AI-powered legal research tool constituted fair use?
Whether the commercial and competitive nature of ROSS’s product affected the fair-use analysis?
Whether ROSS’s use was sufficiently transformative to justify protection under the fair-use doctrine?
Whether the use of Westlaw’s protected material caused or threatened market harm to Thomson Reuters?
Arguments of the Parties
Arguments of Thomson Reuters
Thomson Reuters argued that its Westlaw headnotes were original editorial works and therefore protected by copyright. Although the underlying judicial opinions themselves were not protected in the same manner, Thomson Reuters maintained that its editorial selection, organisation and expression in the headnotes involved sufficient originality.
It was further argued that ROSS had deliberately obtained and used the copyrighted material to create a competing legal research product after Thomson Reuters had refused to license the material. According to Thomson Reuters, this was a commercial use that directly threatened Westlaw’s market position.
Arguments of ROSS Intelligence
ROSS argued that its use of the material should qualify as fair use. It maintained that the material was used to develop a new technological tool rather than to reproduce Westlaw’s headnotes for customers. ROSS also disputed the scope of Thomson Reuters’ copyright claims and argued that copyright law should not prevent technological innovation and the development of AI-powered legal research tools. Its position was that using legal information for AI development could constitute a transformative use rather than simple reproduction.
Judgment
On 11 February 2025, the District Court of Delaware granted partial summary judgment in favour of Thomson Reuters. The court held that ROSS had directly infringed copyright with respect to 2,243 Westlaw headnotes identified in the proceedings.
The court also rejected ROSS’s fair-use defence. It concluded that ROSS’s use was commercial and insufficiently transformative in the circumstances of the case. The court also considered the competitive relationship between the ROSS product and Westlaw and found that the use had significant implications for Thomson Reuters’ existing and potential market. Importantly, the decision concerned a legal research/search tool rather than a generative AI system. ROSS’s product did not generate new legal texts in the manner of modern generative AI systems; it was designed to identify relevant judicial opinions in response to legal research questions. Therefore, the judgment must be understood within the specific facts and technology involved in the case.
Reasoning of the court
The court’s reasoning centred on the application of traditional copyright principles to the use of protected material in developing an AI-powered legal research system.
First, the court distinguished between uncopyrightable judicial opinions and the original editorial expression contained in Westlaw’s headnotes. Judicial opinions themselves are public legal materials, but the editorial work involved in selecting, summarising and expressing particular legal principles may contain sufficient originality to receive copyright protection. The court therefore recognised copyright protection in the relevant Westlaw headnotes.
Second, the court considered whether ROSS had actually copied protected expression. It concluded that the evidence established direct copying of the protected headnotes and identified 2,243 headnotes for which infringement was established. The major issue was the fair-use defence. Under United States copyright law, fair use requires consideration of four principal factors: the purpose and character of the use, the nature of the copyrighted work, the amount used and the effect upon the potential market.
Regarding the purpose and character of the use, the court considered ROSS’s use to be commercial. ROSS was developing a product that competed directly with Westlaw. The court also rejected the argument that the use was sufficiently transformative. Unlike cases where copyrighted material is used for a substantially different expressive purpose, the court considered ROSS’s use closely connected to the creation of a competing legal research product.
With respect to the nature of the copyrighted work, the court recognised that Westlaw’s headnotes were factual and functional to some extent, which weighed in ROSS’s favour. However, the headnotes also contained original editorial expression, giving them meaningful copyright protection. The court also considered the amount and substantiality of the copying. ROSS had used a large body of Westlaw’s protected material to construct its training data. Although the final ROSS product did not simply reproduce the headnotes to customers, the copying necessary to develop the competing system remained significant.
Finally, the court placed substantial importance upon market effects. ROSS’s product competed with Westlaw in the same legal research market. The court therefore considered the potential harm to Thomson Reuters’ market and licensing opportunities significant. The combination of commercial purpose, lack of sufficient transformation and market competition ultimately led the court to reject fair use.
Significance of the Judgment
The decision in Thomson Reuters v. ROSS Intelligence is significant because it represents an important judicial examination of copyright law in the context of AI training. It demonstrates that the use of copyrighted material for developing an AI system does not automatically qualify as fair use. The purpose for which the material is used, the nature of the copyrighted work, the extent of copying and the effect on the copyright owner’s market remain important considerations.
The judgment is particularly important for the legal-technology industry because ROSS was developing an AI-based legal research competitor to Westlaw. The court’s emphasis on commercial competition and market harm demonstrates that courts may scrutinise AI training practices where the resulting product directly competes with the copyright owner’s services. At the same time, the decision should not be interpreted as establishing that all AI training using copyrighted material is unlawful. Its reasoning was based upon the specific facts, the particular copyrighted works involved and the nature of ROSS’s product. Moreover, ROSS’s system was a legal research tool rather than a generative AI model. The case therefore provides an important framework for analysing the relationship between copyright, AI innovation, commercial competition and fair use, while leaving broader questions concerning generative AI training for future litigation.
Conclusion
Thomson Reuters Enterprise Centre GmbH & West Publishing Corp. v. ROSS Intelligence Inc. represents an important development in the emerging field of AI and intellectual property law. The decision demonstrates that technological innovation does not automatically override existing copyright protections. Where protected editorial content is copied to develop a competing commercial AI product, courts may closely examine whether the use satisfies the requirements of fair use. The judgment is particularly relevant to developers of AI systems because it highlights the importance of considering the source, nature and purpose of training data. At the same time, its application must remain fact-specific, especially because ROSS’s technology was a legal research tool rather than a generative AI system. The case therefore provides an important foundation for understanding how copyright principles may apply to AI training while leaving many broader questions for future courts and legislators.
References
U.S. Copyright Act, 17 U.S.C. § 107 — Fair Use Doctrine.
Feist Publications, Inc. v. Rural Telephone Service Co., 499 U.S. 340 (1991).
Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith, 598 U.S. 508 (2023).
Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., District Court Memorandum Opinion, 11 February 2025.
