The Third Circuit has delivered one of the most important appellate rulings yet at the intersection of copyright law, legal publishing, and artificial intelligence. In affirming Thomson Reuters’ win against Ross Intelligence, the court left in place a ruling that Ross infringed Westlaw headnotes and could not shield its conduct under fair use.
For legal professionals, the decision is significant well beyond this long-running dispute. At a high level, it reinforces that editorial content built on top of public-law materials may still receive meaningful copyright protection. While judicial opinions themselves are not copyrightable, the way a legal publisher organizes, summarizes, and adds editorial analysis to that body of law can be. That distinction has always mattered in the legal-research market; it now matters even more in the AI era.
The litigation has been closely watched because Ross’s alleged use of Westlaw headnotes was tied to building an AI-driven legal research product. The appellate ruling therefore lands squarely in the broader debate over whether copyrighted materials can be used to train, improve, or structure AI tools without a license. By refusing to extend fair use here, the court gave content owners a stronger position in arguing that commercial AI development does not automatically transform protected works into permissible inputs.
The underlying Delaware case, Thomson Reuters Enterprise Centre GmbH et al v. ROSS Intelligence Inc., has served as a bellwether for how courts may treat “non-generative” AI uses of copyrighted works. Now, with the Third Circuit affirmance in Thomson Reuters Enterprise Centre GmbH, et al v. Ross Intelligence Inc, litigators and in-house teams have appellate guidance to cite when assessing risk around training data, product development, and licensing strategy.
For law firms and legal departments, the practical takeaway is straightforward: AI procurement and development decisions increasingly require careful rights analysis, especially where a tool relies on third-party editorial content rather than raw public-domain materials. Compliance teams should pay close attention to vendor representations about training data and content provenance. In-house counsel evaluating AI partnerships or internal tools will want to revisit contract language, indemnity provisions, and usage restrictions in light of this ruling.
The opinion is also likely to shape competitive dynamics in legal research. Established publishers may view it as validation of the value of their proprietary enhancements, while AI entrants may face greater pressure to license content or redesign workflows around clearly noninfringing inputs. Either way, this is a decision that will be cited far beyond the legal-research industry.