Third Circuit: Using Westlaw’s Editorial Headnotes To Train AI Is Not Fair Use — What Developers Need To Know

The U.S. Court of Appeals for the Third Circuit has upheld a ruling that Ross Intelligence’s use of Westlaw editorial headnotes to train a competing AI legal-research platform did not qualify as fair use. The decision, which affirms the lower court, sends a clear message to AI developers that the origin and function of training data matter legally — not just what appears in the final product delivered to users.

Public Opinions vs. Proprietary Headnotes

At the center of the dispute is a straightforward distinction: judicial opinions are public-domain material and may generally be copied or repurposed, but the editorial headnotes Westlaw’s editors write to summarize and classify those opinions are proprietary. Headnotes act like a company-built index on top of public records: the underlying data may be free, but the editorial selection, emphasis, and phrasing are original work.

Why the Court Found Copyright Protection

The Third Circuit concluded that Westlaw’s editorial choices — selecting which legal issues to highlight, choosing what factual and analytical context to include, and compressing points for quick retrieval — supplied the "creative spark" copyright law protects. The court applied copyright’s low originality threshold and found that the headnotes embodied sufficient independent expression.

How Ross Used the Headnotes

Ross Intelligence, in partnership with LegalEase Solutions, created training materials linking legal queries to passages in judicial opinions. Those materials incorporated Westlaw headnotes and the court emphasized that the headnotes were functionally central to Ross’s product rather than incidental.

"The undisputed evidence reflects that Thomson Reuters' materials possess a creative spark, and ROSS aspired to be a direct competitor by using them for a highly similar purpose," Judge Tamika Montgomery-Reeves wrote for the court.

Transformative Use And Market Harm

The court rejected Ross’s argument that internal, non-user-facing use of the headnotes was transformative. At best, the use was "minimally transformative": although Ross trained models with the material rather than presenting verbatim headnotes to users, the commercial purpose — helping users perform legal research — was highly similar to Westlaw’s service. The court also found likely market harm to Westlaw’s existing research business and to a potential licensing market in which Thomson Reuters might sell headnotes to AI developers.

Scope And Practical Takeaway

The Third Circuit stressed that its decision is fact-specific. Key factors included protectable editorial content, direct commercial competition, copying of a substantial portion of the headnotes, and a product aimed at the same market function. The ruling does not categorically prohibit training AI on copyrighted works, but it does signal real legal risk when proprietary summaries, labels, annotations, or metadata are used to train competing systems.

Digital-rights groups such as the Electronic Frontier Foundation warned that the decision could give rights holders broad control over materials that functionally support products, a concern that will likely appear in future cases over training datasets. Ross Intelligence is now defunct, but the opinion will be closely watched by developers working with legal databases, news archives, annotated scientific literature, and other structured sources.

Bottom line: Courts will examine not just whether material is publicly available, but whether the training content embodies original editorial judgment and enabled a competing commercial product — factors that can weigh strongly against fair use.