Introduction
In a landmark legal decision, Thomson Reuters secured a significant victory against ROSS Intelligence, setting crucial precedents in the realm of AI-related copyright litigation. The case centered on the unauthorized use of Thomson Reuters’ Westlaw content by ROSS to train its AI model, raising important questions about fair use in the context of artificial intelligence.
Description
In a pivotal legal victory for Thomson Reuters, a recent court ruling in the case against ROSS Intelligence established important precedents regarding fair use in AI-related copyright litigation [6]. The court determined that ROSS [1] [3] [6] [7], a now-defunct legal research firm [4], unlawfully utilized content from Thomson Reuters’ Westlaw platform to train its AI model without authorization [4]. Thomson Reuters initiated the lawsuit in 2020 [5], claiming that ROSS’s actions constituted copyright infringement. Judge Stephanos Bibas of the 3rd US Circuit Court of Appeals found that ROSS directly infringed on Thomson Reuters’ copyright concerning the use of case law headnotes organized by Westlaw’s proprietary Key Number system. The ruling highlighted actual copying and substantial similarity [7], with at least 2,243 of the 2,830 Bulk Memos created by ROSS being copied and substantially similar to Thomson Reuters’ original work, providing strong circumstantial evidence of infringement. The court rejected ROSS’s defenses related to innocent infringement [6], copyright misuse [1] [3] [7], merger [3] [4] [8], and scènes à faire [3].
The ruling emphasized the commercial nature of ROSS’s use [6], which weighed heavily against its fair use defense [6]. The court concluded that ROSS’s use was not transformative, as it did not serve a different purpose than Thomson Reuters’ original use [7]. Judge Bibas noted that if ROSS’s AI had only learned language patterns to generate new content [2], it could have been considered transformative; however [2], since it directly reproduced the headnotes [2], it did not qualify for fair use [2]. The court’s determination that ROSS intended to create a direct competitor to Thomson Reuters further diminished the original value of the copyrighted material, reinforcing the ruling against fair use.
The court’s analysis of the four fair use factors revealed a split decision [1]. The first factor [1], assessing the purpose and character of the use [1], favored Thomson Reuters due to ROSS’s for-profit nature and the non-transformative use of the copyrighted material [1]. The fourth factor [1] [6], which evaluates the effect on the market for the copyrighted work [1], heavily favored Thomson Reuters [1] [5] [8]. The court recognized ROSS’s AI tool as a potential market substitute for Westlaw, raising significant fair use concerns and highlighting its implications for the licensing market for training data for other AI models [1]. The court acknowledged the existence of a potential market for legal AI training data [2], despite ROSS’s claims to the contrary [2]. The only remaining issue for trial is whether some headnotes’ copyrights had expired or were untimely registered [8].
The second and third factors concerning the nature of the copyrighted work and the output of the AI model presented a more nuanced analysis. While the court acknowledged that the headnotes had some originality [1], it deemed them not highly creative. Additionally, the AI model’s output did not reproduce the copyrighted material directly [1], which complicated the assessment of these factors.
This ruling represents a significant victory for content owners in the context of AI-related copyright disputes [3], and the implications of this decision for future cases involving AI and fair use will be closely monitored [3]. Critics have raised concerns that the court’s reasoning undermines potential fair use defenses in AI-related cases, suggesting that the ruling could expand copyright enforcement to intermediate copies and blur the line between fact and expression. The court’s application of the merger doctrine [8], which protects the expression of facts [8], has also faced scrutiny for potentially misapplying established legal principles.
This case is part of a broader trend [4], as numerous lawsuits have emerged from authors [4], visual artists [4], and music labels against AI developers for similar copyright infringement issues [4]. The court’s refusal to recognize fair use for non-generative AI raises concerns for future cases involving large language models and generative outputs [6], suggesting that fair use may be less likely if the AI-generated content is substantially similar to original works and negatively impacts the market for those works [6] [7]. The evolving relationship between AI technology and copyright law [2], particularly in the legal sector [2], underscores the complexities of these issues as they continue to develop. AI developers and deployers are advised to closely monitor ongoing litigation and consider the market implications of using copyrighted materials for training AI models or distributing AI-generated outputs [6], as fair use determinations are highly fact-dependent and may vary across different jurisdictions [6].
Conclusion
The ruling in favor of Thomson Reuters against ROSS Intelligence marks a pivotal moment in AI-related copyright litigation [7], with significant implications for the future of fair use defenses in similar cases. As AI technology continues to evolve, the legal landscape surrounding copyright and AI will require careful navigation, particularly concerning the use of copyrighted materials in training AI models [7]. This decision underscores the importance for AI developers to remain vigilant about the legal ramifications of their practices, as the boundaries of fair use in the context of AI remain complex and highly fact-dependent.
References
[1] https://www.jdsupra.com/legalnews/ross-ai-decision-gives-early-indication-1965012/
[2] https://www.jdsupra.com/legalnews/what-thomson-reuters-v-ross-does-and-2028153/
[3] https://natlawreview.com/article/court-training-ai-model-based-copyrighted-data-not-fair-use-matter-law
[4] https://apnews.com/article/ai-artificial-intelligence-reuters-4a127c5b7e8bb76c84499fe12ad643c8
[5] https://www.computerworld.com/article/3823125/first-judgment-ai-company-loses-copyright-dispute.html
[6] https://www.hunton.com/insights/legal/fair-warning-ais-first-copyright-fair-use-ruling-thomson-reuters-v-ross
[7] https://natlawreview.com/article/fair-warning-ais-first-copyright-fair-use-ruling-thomson-reuters-v-ross
[8] https://www.authorsalliance.org/2025/02/13/thomson-reuters-v-ross-the-first-ai-fair-use-ruling-fails-to-persuade/
