Third Circuit affirms that ROSS's AI training on Westlaw headnotes was not fair use

On September 29, 2026, the US Court of Appeals for the Third Circuit decided Thomson Reuters v. ROSS Intelligence (No. 25-2153), an interlocutory appeal from Judge Stephanos Bibas, who had heard the case in the District of Delaware by designation. Writing for a panel of Judges Restrepo, Montgomery-Reeves and Bove, Judge Montgomery-Reeves affirmed partial summary judgment for Thomson Reuters: the 2,243 Westlaw headnotes at issue are original enough for copyright protection, and ROSS’s use of them was not fair. “Under ROSS’s framing, this case appears to concern the future of AI legal technology,” the court wrote. “In truth, this is no more than an ordinary copyright case.”

ROSS had hired LegalEase to write about 25,000 training memos, each a legal question with four to six responsive passages, and the memo writers used Westlaw headnotes to build them. The court found ROSS’s use “minimally transformative, at best”, because both companies use the headnotes to build a legal research platform. The second factor leaned slightly toward fair use, but the first, third and fourth weighed against it, the fourth because ROSS was building a commercial substitute for Westlaw.

The court went out of its way to say ROSS’s AI “was not a generative AI”: it only returned passages from existing judicial opinions. A footnote addresses the Justice Department’s September 1, 2026 statement of interest in the OpenAI copyright litigation, which relied on Bartz v. Anthropic to argue that training a large language model that can “generate original responses” is transformative. Those concerns, the court said, “do not apply here.”

Why it matters: it moves the training-data fight up from the district courts to a federal court of appeals, binding precedent across the Third Circuit, and the answer went against the AI developer. What it does not show: the panel deliberately limited the holding to a non-generative system built to compete directly with its source, so it does not decide the generative-model cases against OpenAI, Meta and others. It is also an interlocutory ruling; damages and the rest of the case go back to the district court.

Sources

Last verified October 5, 2026