Copyright law’s evolving relationship with artificial intelligence persists as an unresolved legal question, particularly as courts examine its boundaries. On February 11, 2025, Judge Stephanos Bibas of the U.S. Court of Appeals for the Third Circuit delivered a ruling in Thomson Reuters Enter. Ctr. GmbH v. Ross Intel. Inc., focusing on whether AI training data qualifies as fair use under copyright doctrine.
The court ruled that Ross Intelligence’s use of Thomson Reuters’s Westlaw headnotes to train a nongenerative AI tool was not protected by fair use. However, the ruling is not the final word on the matter, as the Third Circuit granted Ross permission to pursue an interlocutory appeal, which was heard on June 11, 2026.
Despite the significance of the Thomson Reuters v. Ross ruling, its applicability has been narrower than some headlines suggested. Key differences, such as the nature of Ross’s product, access to the copyrighted works, and the type of AI involved, limited the breadth of the case’s influence.
Competing Products and Fair Use
A central reason Ross lost on fair use’s fourth factor was that its AI tool directly competed with Thomson Reuters’s Westlaw platform. Ross used Westlaw’s copyrighted headnotes to develop a similar legal research product, creating a clear market substitute.
In contrast, many AI copyright cases involve generative AI systems trained on publicly available online articles, where the content owners aren’t developing competing AI tools. Without direct market competition, courts may reach different conclusions on fair use’s market-effect factor.
Paywalls and Access Restrictions
Westlaw’s paywall played a significant role in the court’s decision, as it signaled Thomson Reuters’s intent to control how the headnotes were used. Ross’s unauthorized access and use of this content further undermined its fair use argument.
This contrasts with other AI lawsuits that involve freely accessible online materials. When content is publicly available, especially without paywalls or usage restrictions, courts might weigh fair use factors differently, particularly if the AI tool’s use is more transformative.
The distinction between generative and nongenerative AI is also critical, as the court’s decision was influenced by the type of AI at issue. Ross’s nongenerative AI tool used the headnotes to enhance a legal research tool that performed the same basic function as Westlaw, whereas generative AI models synthesize training data into entirely new outputs.
In Bartz v. Anthropic and Kadrey v. Meta, two Northern District of California judges ruled on closely related questions involving large language models trained on copyrighted books. The courts found that training these models on lawfully acquired copyrighted works was fair use, but that using pirated works was infringing.
The Bartz case settled for $1.5 billion, the largest copyright class action settlement on record, while Kadrey remains active. These rulings highlight the importance of lawful acquisition of training data and the need for companies to vet the provenance of their data.
The distinction between lawful acquisition and unlawful acquisition of training data is critical, as seen in Bartz v. Anthropic and Kadrey v. Meta. Companies that train generative models on copyrighted works they never had the right to possess may find fair use unavailable, regardless of how transformative the resulting model is.
Leave a Reply