Third Circuit affirms Thomson Reuters on Westlaw headnotes

On 29 September 2026 the U.S. Court of Appeals for the Third Circuit filed its opinion in Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., No. 25-2153. Judge Montgomery-Reeves wrote for a panel of Judges Restrepo, Montgomery-Reeves, and Bove. Argument was on 11 June 2026. The panel affirmed a grant of partial summary judgment for Thomson Reuters.

The appeal is from the District of Delaware, Judge Stephanos Bibas sitting by designation, No. 1:20-cv-00613. The opinion says the district court granted partial summary judgment and certified two questions under 28 U.S.C. § 1292(b): whether the West headnotes and Key Number System are original as a matter of law, and whether ROSS’s use of the headnotes was fair use. The opinion does not print the date of that order. Haynes Boone says Judge Bibas granted it on 11 February 2025. That date is the client alert’s.

What the panel held

The opinion says the headnotes are copyrightable and that ROSS’s use was not fair, and it will affirm partial summary judgment. The judgment covers 2,243 Westlaw headnotes. A footnote says the district court reached that number from a batch of 2,830, where memo questions resembled the headnote text and not the underlying opinions.

Editors chose which points of law to include and how to word them. The opinion’s phrase for the creativity bar is the “creative spark.”

ROSS answered plain-language questions with passages from opinions. The opinion states: “ROSS’s AI was not a generative AI, meaning it would not create any new expression; it would only return text passages from preexisting judicial opinions.” It could draw on about ten million uncopyrighted opinions. LegalEase wrote about 25,000 training memos, using thousands of Westlaw headnotes to frame the questions. For this appeal, ROSS is not challenging that LegalEase’s and Morae Global’s copying is attributable to ROSS.

Fair use

ROSS aimed to sell a legal-research platform at prices comparable to Westlaw. “So ROSS’s use is commercial.” Thomson Reuters uses headnotes so researchers can find opinions. ROSS used them to train a platform with the same job. The training step, the opinion says, “arguably presents a slight degree of difference in use.” The uses still “share the same ultimate purpose,” “making ROSS’s use minimally transformative, at best.” The opinion says this use “aims to replace Westlaw,” unlike the search feature in Authors Guild v. Google. On intermediate copying, the opinions were already available. The sentence in the opinion is: “Unlike necessity, ease is not a justification for copying.”

The second factor weighs slightly in favor of fair use. The headnotes were published, and they are “more factual than fictional.”

On the third factor, the opinion says ROSS copied the entire text of the 25,000 headnotes in the memos. ROSS called that 0.08 percent of 28 million headnotes. The panel says that was more than necessary. The 2,243 figure is the judgment. The 25,000 figure is the training set discussed in this factor.

On the fourth factor, the opinion finds harm in the legal-research market, because ROSS built a competing platform, and in a derivative market. It states: “Here, the evidence shows that the market for licensing headnotes as text to train AI is rapidly developing.” Thomson Reuters also uses its headnotes to train its own AI search products. The opinion says the copying “usurped Thomson Reuters’s opportunity to enter that derivative market and license its headnotes for that purpose.” It rejects a public-benefit defense: the opinions are free, the service was priced like Westlaw, and ROSS showed no evidence that the ruling will halt AI development.

“In conclusion, the second factor weighs slightly in favor of fair use, but the first, third, and fourth factors weigh against it.” The panel holds the use was not fair, and it will affirm.

Where the opinion stops

Judge Montgomery-Reeves writes that the case can look like a fight over the future of AI legal technology. “In truth, this is no more than an ordinary copyright case.” A footnote discusses a Justice Department statement about generative AI in the OpenAI multidistrict litigation in the Southern District of New York, and it cites Bartz v. Anthropic and Kadrey v. Meta. “The concerns raised in that separate case do not apply here.” ROSS’s platform “cannot generate original expression,” and it was trained as a commercial substitute for Westlaw. The opinion quotes the Department’s phrase for the other technology, models that can “generate original responses,” and sets that record aside.

Haynes Boone calls the decision what “appears to be the first federal appellate ruling on fair use in AI training.” That line is the firm’s. The opinion does not use it.

The holding a reader can rely on is partial summary judgment: 2,243 headnotes are original, and this non-generative competitor’s use of them was not fair. The footnote is the panel’s own warning against reading the case as a decision about generative models.