Thomson Reuters v. Ross, explained

Thomson Reuters sued Ross Intelligence for training a legal-research tool on 2,243 Westlaw headnotes. In February 2025, Judge Bibas in the District of Delaware granted summary judgment rejecting Ross’s fair-use defense — the first US ruling to do so for AI training.

Why it matters to a buyer

The ruling is often cited as if it condemns all AI training. It does not: Ross built a non-generative search tool that substituted directly for Westlaw, and the court’s market-harm reasoning turned on that. Read it as a warning about training a direct substitute on a rival’s licensed content, not a blanket rule for generative LLMs — and watch the appeal, because the appellate answer will bind more broadly.

Why it matters to a data owner

The case shows a court protecting an owner’s licensing market when a trained product competes with the source. That market-harm logic is the strongest argument licensed suppliers have.

Current legal status

February 2025: summary judgment for Thomson Reuters in the District of Delaware (Judge Bibas), rejecting fair use for training on 2,243 Westlaw headnotes — a non-generative case, so its reach to generative models should not be overstated. It is now the first AI-training fair-use case at the appellate level: the Third Circuit heard oral argument on June 11, 2026, before Judges Restrepo, Montgomery-Reeves, and Bove. Questioning centered on factor one (whether the use was transformative) and factor four (market harm), against the backdrop of the circuit’s recent fair-use decision in ASTM v. UpCodes. As of July 21, 2026, the decision is pending.

What fiund does about it

However the appeal lands, licensed data does not depend on it. fiund’s licences grant training rights explicitly, in writing.

Sources

• LawNext — Third Circuit presses Ross and Thomson Reuters on fair use and market harm
• Baker Botts — Third Circuit hears oral argument (July 2026)

← All rights & provenance guides