Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc. — (3d Cir. 2026)

Thomson Reuters v. ROSS Intelligence, decided by the Third Circuit on September 29th, 2026. In an opinion by Judge Montgomery-Reeves, the court held that Westlaw's headnotes are original enough for copyright protection, and that ROSS's use of them to train a competing, non-generative AI legal-research tool was not fair use. The court affirmed the district court's partial summary judgment, which this series briefed in 2025. Here's the brief.

Thomson Reuters owns Westlaw. Its editors write headnotes, short statements of the points of law in a judicial opinion, under detailed drafting guidelines. Thomson Reuters disclaims any copyright in the opinions themselves.

ROSS set out to build an AI search engine that would answer plain-language legal questions with passages from judicial opinions. Its AI was not generative; it returned only text from existing opinions. To train it, ROSS commissioned about twenty-five thousand memos, each posing a legal question with responsive passages. The memo writers used Westlaw headnotes to frame the questions. ROSS marketed its platform as a Westlaw competitor, at comparable prices.

Thomson Reuters sued. Judge Bibas, a circuit judge sitting by designation in Delaware, granted partial summary judgment that ROSS's use of 2,243 headnotes infringed and was not fair use. He then certified two questions for interlocutory appeal: whether the headnotes are original, and whether ROSS's use was fair.

Originality is an extremely low bar under the Supreme Court's decision in Feist: a work needs only independent creation and a modicum of creativity — some creative spark.

Fair use turns on the four statutory factors, and ROSS, as the proponent of the defense, bore the burden on each.

On originality, the court held that each of the headnotes has the required creative spark. Editors chose which points of law mattered and how to state them so each headnote stands on its own. The court found support in the Supreme Court's 1888 decision in Callaghan v. Myers. And it rejected ROSS's argument that the copyright would give Thomson Reuters a monopoly over the law.

On fair use, the first factor weighed against ROSS. Its use was commercial, and training an AI was only an intermediate step toward the same goal Westlaw serves: helping users find responsive legal material.

The court distinguished the generative-AI decisions in Bartz v. Anthropic and Kadrey v. Meta, noting that ROSS's platform cannot generate original expression. It also distinguished Google v. Oracle and the intermediate-copying cases, where copying was necessary to reach unprotected functional elements. ROSS could have used the freely available opinions.

The second factor favored ROSS, slightly, because headnotes are more factual than fictional. The third weighed against it: ROSS took more than necessary. The fourth weighed against it as well. ROSS's copying diminished the headnotes' value as a draw to Westlaw, and usurped a developing market for licensing headnotes as AI training data. The court found no evidence for ROSS's claimed public benefits.

Thomson Reuters v. ROSS is a federal appellate decision applying fair use to the training of an AI system. It treats the question as an ordinary application of the four factors, governed by Warhol's focus on whether the copier's purpose differs from the original's. The decision is expressly confined to a non-generative tool built to substitute for the original. It distinguishes, rather than resolves, the question whether training generative models is transformative.

Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., Third Circuit No. 25-2153, decided September 29th, 2026. I'm John Goodhue. Thanks for watching.