On February 11, 2025, the U.S. District Court for the District of Delaware delivered a significant ruling in the case of Thomson Reuters Enterprise Centre GmbH v. Ross Intelligence Inc.[1], favoring Thomson Reuters and setting a precedent regarding the use of copyrighted works in training AI models. The court determined that such use does not qualify for the ‘fair use’ exception under copyright law, which has substantial implications for both AI developers and copyright holders. You can expect this case will be used as persuasive authority in the currently pending music industry lawsuits filed by the Recording Industry Association of America (“RIAA”) against AI music generators Suno AI and Uncharted Labs, Inc.
Background of the Case
Thomson Reuters, the owner of Westlaw—a leading legal research platform—provides not only access to legal judgments and statutes but also offers editorial content like headnotes that summarize key case aspects. Westlaw employs a proprietary classification system known as the Key Number System, organizing legal content methodically.
In contrast, Ross Intelligence developed a competing platform that utilizes artificial intelligence to enhance legal research. When Ross sought permission to use Westlaw’s content for training its AI, Thomson Reuters denied the request, seeing it as a conflict of interest. As a workaround, Ross worked with LegalEase to obtain ‘Bulk Memos’, which were compilations of legal questions and answers, some of which were derived from Westlaw headnotes without direct reproduction. This approach enabled Ross to train its AI tool, ultimately leading Thomson Reuters to file a lawsuit in May 2020, claiming copyright infringement.
The Ruling and Fair Use Evaluation
Initially, in 2023, the court did not grant Thomson Reuters a summary judgment on most claims. However, following a thorough reevaluation, the court invited both parties to submit renewed pleadings on the matter. On February 11, 2025, Judge Stephanos Bibas ruled in favor of Thomson Reuters, granting them partial summary judgment for direct copyright infringement and dismissing Ross’s fair use claims.
In its analysis, the court examined four critical factors concerning the fair use doctrine:
- Purpose and Character of Use: The court affirmed that Ross’s intentions were commercial, as it aimed to develop a competitive product, rather than adding new meaning or character to the original work. Moreover, Ross’s claim of intermediate copying as a protective measure was rejected, as the court determined that the type of copying involved wasn’t necessary for innovation.
- Nature of the Copyrighted Work: While there was some originality in Thomson Reuters’ work, it did not feature a high enough level of creativity to affect the outcome positively for Ross.
- Amount of Work Used: Here, Ross benefitted as the headnotes were not copied into the final product. However, the court noted that this factor alone did not negate the infringement.
- Effect on Market Value: Ultimately, this factor weighed heavily in favor of Thomson Reuters. The court found that Ross’s platform was a direct competitor to Westlaw, potentially hindering its market position.
Despite both parties having points in their favor, the court inclined towards Thomson Reuters, emphasizing the significant impact of the fourth factor, consistent with prior Supreme Court precedents. Consequently, it concluded that fair use cannot apply in cases where copyrighted material is used to train AI without explicit permission.
Implications for the Future
This ruling is pivotal for copyright holders. It reinforces their rights over proprietary works and suggests that licensing arrangements could become critical in the landscape of AI development. Consequently, AI providers must now tread carefully, ensuring they utilize either non-copyrighted materials or secure necessary licenses to mitigate legal risks.
While this decision solidifies restrictions around copyright and AI training, it should be noted that it does not set an absolute precedent. Judge Bibas clarified that Ross’s research platform does not constitute a generative AI tool, which opens a dialogue regarding how future cases surrounding generative AI might unfold, including the currently pending music industry cases brought by the RIAA in June 2024 against AI Music generators Suno AI and Uncharted Labs, Inc. Our previous deep dive article on those cases is linked here. As I discussed on Wall Street Week, AI generators could avoid such issues of infringement by simply licensing the copyrighted material they wish to train their AI. As the legal landscape continues to evolve, companies in the AI sector and those impacted by it (including the entertainment and music industry) would benefit from closely monitoring these developments to navigate the challenges of copyright laws effectively while fostering innovation and protecting creators’ rights.
Jenifer Wallis is the Chair of Munck Wilson Mandala’s Entertainment Practice and is located in MWM’s Los Angeles office.
[1] 2025 WL 458520 (D. Del. Feb. 11, 2025).
