Why AI Companies Should Be Worrying About Thomson Reuters v. Ross

In a major ruling on February 11th, the U.S. District Court for the District of Delaware, presided over by Judge Stephanos Bibas, addressed key issues at the intersection of artificial intelligence (AI) and copyright law in Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc.

Ronin Legal has been closely following the developments in this evolving legal landscape, including OpenAI’s efforts to consolidate the copyright lawsuits against them and the broader implications of Raw Story Media & AlterNet Media Inc. v. OpenAI. With this decision in the Reuter’s case, the courts are making further headway in defining the boundaries of copyright protection in the age of AI.

OVERVIEW

Thomson Reuters (“Reuters”), the parent company of Westlaw, sued Ross Intelligence (“Ross”) in 2020, alleging that Ross unlawfully copied and used Westlaw’s copyrighted content – such as case summaries and headnotes – to train its AI-powered legal research tool. The primary legal issue here revolved around whether Ross’s use of the material to develop an AI-driven research tool amounted to copyright infringement or if it could be justified under the “fair use doctrine.”

KEY COPYRIGHT ISSUES INVOLVED

Reuters’ main contention was that Westlaw’s summaries and headnotes were creative expressions by legal experts, not mere factual compilations, and therefore protected under copyright law. The company contented that Ross’ actions amounted to unauthorized reproduction, copyright infringement, and directly harmed Westlaw’s market by creating a competing product. Anticipating Ross’ fair use defence, Reuters also emphasized the commercial nature of their use and asserted that it was not transformative in nature. Rather than adding new expression or meaning, Reuters claimed Ross’ platform functioned as a market substitute, replicating Westlaw’s core functionalities without authorization.

Ross, on the other hand, argued that its AI development qualified as fair use, asserting that its use of Westlaw content was transformative – a defense frequently invoked in similar cases and extensively discussed by the court in Authors Guild v. OpenAI. While acknowledging the commercial nature of its use, Ross contended that it remained transformative because the headnotes were not simply reproduced but instead converted into numerical data representing relationships among legal terms within its AI system. The company argued that this transformation created a new and innovative tool, rather than serving as a mere substitute for Westlaw’s original content.

Ross further argued that Westlaw’s headnotes were merely factual summaries of judicial opinions, affording them a lower degree of copyright protection and strengthening its fair use claim. It contended that its copying was limited and necessary to achieve a transformative purpose, emphasizing that its AI did not reproduce the headnotes verbatim but instead analysed them to enhance its understanding, enabling it to generate original outputs based on judicial opinions.

THE COURT’S RULING

The court granted partial summary judgment to Reuters on direct copyright infringement, fair use, and other defenses, while denying the summary judgment motions from Ross. In its decision, the court analyzed the four fair use factors outlined in the U.S. Copyright Act (17 U.S.C. § 107):

  1. The purpose and character of use, including whether the use is of a commercial nature or for nonprofit educational purposes;

The court determined that Ross’s use was commercial and non-transformative. While Ross claimed its AI analyzed the headnotes to learn language patterns, the court found that Ross’s platform aimed to compete directly with Westlaw by offering a similar legal research service. The use did not add new expression or meaning but sought to replicate Westlaw’s functionalities, thereby failing the transformative use test.

  1. The nature of the copyrighted work;

The court ruled in Ross’ favour on this fair use factor, finding that while Westlaw’s content meets the minimal threshold for originality, it is not highly creative. Although the headnotes contain some editorial input, the “Key Number System” (a legal research tool developed by Westlaw to organize and categorize case law) is primarily a factual compilation with limited creative expression.

  1. The amount and substantiality of the portion used in relation to the copyrighted work as a whole;  

Ross had copied a substantial number of headnotes – over 2,000 – for its AI training. Despite this high number of headnotes used, the court ruled in Ross’ favour on this fair use factor as well. It emphasized that the main issue was not how much content Ross copied, but rather how much of that content was made publicly accessible and whether it served as a market substitute for Westlaw. Since Ross’ AI did not reproduce or display Westlaw’s headnotes in its output, the court found no factual dispute on this point, supporting Ross’ fair use argument in this regard.

  1. The likely effect of the use on the potential market for the copyrighted work;  

The court found that this fair use factor was in favour of Reuters, emphasizing that Ross could have developed its product without infringing Westlaw’s copyrights. In assessing the potential effect of Ross’ copying on Westlaw’s market, the court initially considered whether Ross’ AI tool served a different purpose by creating a new research platform. However, it ultimately concluded that Ross intended to compete directly with Westlaw and usurp its market share, and that the company had failed to prove otherwise.

Courts have long debated whether the first or fourth fair use factor is the most significant, but in this case and context, the court highlighted the importance of the fourth factor, as both it and the first factor weighed against Ross. In summary, the court held that Ross’s actions infringed upon Reuters’ copyrights and that the fair use defence was inapplicable in this context.

WHY SHOULD THE AI INDUSTRY CARE?

This ruling is a milestone in defining AI’s relationship with copyright law, particularly in the legal and publishing industries. It implies that AI developers cannot assume that scraping or repurposing copyrighted text for training purposes will automatically qualify as fair use, and that such actions must still align with established copyright laws. The court’s emphasis on the non-transformative nature of Ross’s use and the potential market harm to Reuters also reflects the judiciary’s inclination to protect the rights of original content creators against unlicensed commercial exploitation.

Moreover, the decision provides a critical reference point for future disputes at the intersection of AI development and copyright infringement. It signals to AI developers the importance of securing appropriate licenses for copyrighted materials and cautions against assuming that fair use defences will prevail in cases where the new use does not sufficiently transform the original work or where it competes directly with the source material’s market.

With AI technology advancing faster than legal frameworks can adapt, 2025 could be a defining year for AI copyright law. Ronin Legal will continue to monitor these developments and provide insights into their impact on the industry.

 

Authors:  Shantanu Mukherjee, Varun Alase