NextFin

ROSS Cites Anthropic Fair Use Case in 3rd Circuit IP Appeal Against Westlaw, March 2026

Summarized by NextFin AI
  • ROSS Intelligence is appealing a 2025 ruling that found it infringed on Westlaw’s copyrighted headnotes, arguing that recent developments in the Concord Music Group v. Anthropic case support its fair use defense.
  • The distinction between 'input' and 'output' is central to the appeal, with ROSS claiming its AI uses copyrighted material as a structural scaffold rather than replacing it.
  • The outcome of the case could impact the AI industry significantly, potentially establishing a 'pay-to-train' model that restricts access to proprietary datasets.
  • Legal analysts believe the Third Circuit's decision will hinge on the effect of ROSS's use on the potential market for Westlaw's copyrighted work, weighing transformative use against potential licensing revenue loss.

NextFin News - The legal battle over the data that fuels artificial intelligence has reached a critical inflection point as ROSS Intelligence, a defunct legal-tech startup, attempts to revive its defense against Thomson Reuters by leaning on a high-profile copyright dispute involving Anthropic. In a filing with the U.S. Court of Appeals for the Third Circuit on March 13, 2026, ROSS argued that recent developments in the Concord Music Group v. Anthropic case provide a roadmap for why its use of Westlaw headnotes should be protected under the fair use doctrine.

The move is a direct challenge to a 2025 ruling by U.S. District Judge Stephanos Bibas, who found that ROSS had infringed on 2,243 of Westlaw’s copyrighted headnotes to train its natural language processing system. Bibas had previously rejected the fair use defense, characterizing ROSS’s actions as commercial and non-transformative. However, ROSS now contends that the legal landscape has shifted, suggesting that the mere act of using copyrighted material as an intermediate "training input" does not constitute the same market harm as displaying that content to end users.

At the heart of the appeal is the distinction between "input" and "output." In the Anthropic case, the court has been forced to grapple with whether an AI model’s internal processing of data—even if that data is copyrighted—is a transformative use that serves a different purpose than the original work. ROSS argues that its AI did not "replace" Westlaw’s headnotes; rather, it used them as a structural scaffold to help its software understand the relationship between legal concepts. The headnotes themselves were never shown to ROSS users, a fact the company claims aligns with the "intermediate copying" precedents that have historically protected search engines and software interoperability.

The stakes for the broader AI industry are immense. If the Third Circuit upholds the original ruling, it would solidify a "pay-to-train" model that could effectively gatekeep the development of specialized AI tools behind the proprietary datasets of legacy incumbents. Thomson Reuters, which owns Westlaw, has argued that allowing companies like ROSS to scrape its curated headnotes for free would destroy the incentive to invest in high-quality legal editorial work. They maintain that the "potential market" for AI training data is a legitimate property right that must be protected from unlicensed exploitation.

The tension between these two positions reflects a generational divide in the legal-tech sector. For decades, Westlaw and LexisNexis have maintained a duopoly on legal research through massive, human-curated databases. ROSS represented the first wave of AI challengers seeking to break that grip by using machine learning to automate the synthesis of case law. By citing the Anthropic case, ROSS is attempting to link its struggle to the wider Silicon Valley narrative, where the "fair use" of the open web is seen as a prerequisite for technological progress.

Legal analysts suggest that the Third Circuit’s decision will likely hinge on the fourth factor of the fair use test: the effect of the use upon the potential market for the copyrighted work. While Judge Bibas found that ROSS’s use could harm Thomson Reuters’s ability to license its data for AI training, ROSS is now betting that the appellate court will view the "transformative" nature of AI training as more significant than the theoretical loss of licensing revenue. The outcome will determine whether the next generation of legal AI is built on open innovation or through the expensive toll booths of the industry's established giants.

Explore more exclusive insights at nextfin.ai.

Insights

What is fair use in the context of artificial intelligence?

What are the key developments in the Concord Music Group v. Anthropic case?

What led ROSS Intelligence to cite the Anthropic case in its appeal?

How does the current market landscape affect ROSS's appeal against Westlaw?

What are the implications of the Third Circuit's ruling for the AI industry?

What changes occurred in copyright law relevant to AI since the 2025 ruling?

What are the potential consequences for AI development if ROSS loses its appeal?

What arguments does Thomson Reuters present against ROSS's use of its headnotes?

How does ROSS differentiate its use of Westlaw's headnotes from traditional copying?

What historical precedents support ROSS's fair use argument?

How has the legal-tech sector evolved in response to AI challenges?

What are the long-term implications of a 'pay-to-train' model for AI tools?

How does the concept of transformative use apply to AI training?

What factors contribute to the generational divide in the legal-tech sector?

What role does market harm play in the fair use test for AI applications?

What are the differences between ROSS's AI approach and traditional legal research methods?

How might the outcome of the ROSS vs. Westlaw case impact future AI innovations?

Search, ask or analyze anything...