On February 11, 2025, a U.S. federal court issued a landmark ruling on fair use in AI training data. In Thomson Reuters Enter. Ctr. GmbH v. Ross Intel. Inc., the court found that Ross Intelligence’s use of Thomson Reuters’s Westlaw headnotes to train a nongenerative AI tool was not protected by fair use.
At the time of the district court’s ruling, the decision sparked widespread media coverage, with many commentators suggesting it set a precedent for the flood of AI-related copyright lawsuits then pending. However, the ruling has had a narrower applicability than some headlines suggested.
Why the Thomson Reuters v. Ross Ruling Was Limited
Key differences limited the breadth of the case’s influence. These included the nature of Ross’s product, access to the copyrighted works, and the type of AI involved.
Ross’s AI tool directly competed with Thomson Reuters’s Westlaw platform, creating a clear market substitute. This harmed Thomson Reuters’s potential market, including the market for licensing training data.
Westlaw’s headnotes are behind a paywall, available only to subscribers. Ross’s unauthorized access and use of this content undermined its fair use argument.
Generative vs. Nongenerative AI: A Critical Distinction
The type of AI at issue was a critical factor in the court’s decision. Ross’s nongenerative tool used the headnotes to enhance a legal research tool that performed the same basic function as Westlaw. Because the use was not transformative, the court found against Ross on fair use’s first factor.
Many AI copyright cases involve generative AI models that synthesize training data into entirely new outputs. Courts have since addressed how these transformative uses fit into the fair use framework.
Within weeks of the Thomson Reuters decision, two Northern District of California judges ruled on closely related questions involving large language models trained on copyrighted books. In Bartz v. Anthropic, the court held that training Anthropic’s models on lawfully acquired copyrighted books was fair use. However, the court found infringing when books were obtained from pirate libraries.
A similar split was reached in Kadrey v. Meta, involving Meta’s Llama models. These rulings sharpen the distinction between generative and nongenerative AI, as well as lawful and unlawful acquisition of training data.
The line that matters most is not only generative versus nongenerative AI, but also lawful acquisition versus unlawful acquisition of the underlying training data. Companies should vet the provenance of training data, not just how they classify the type of AI involved.
The Ongoing Appeal and Its Implications
The Thomson Reuters v. Ross case is not yet fully resolved. The Third Circuit granted Ross permission to pursue an interlocutory appeal, and oral arguments were heard on June 11, 2026. As of this writing, the appellate decision remains pending, and district court proceedings are stayed.
Reporting from the oral argument suggests that the appellate panel expressed skepticism toward Ross’s transformative-use theory. However, the outcome remains genuinely open, and the Third Circuit’s ruling will be the first federal appellate decision on fair use and AI training data.
Practical Steps for Creators and Businesses
While the legal environment continues to evolve, creators and businesses can take proactive measures to protect their intellectual property. Use strong contracts: Clear licensing terms and usage restrictions can prevent unauthorized use of your content, even for AI training purposes.
