ROSS Intelligence has asked the full US Court of Appeals for the Third Circuit to rehear the first federal appeals ruling on AI training and fair use. Its petition, filed on 3 November, asks the same three judges or the whole court to look again, AI Affairs reports.
The ruling went against ROSS on both questions the court took up: whether the material was protected and whether copying it was fair.
What the court decided
In an opinion dated 29 September 2026, Judge Tamika Montgomery-Reeves wrote for a panel that affirmed a 2025 trial ruling for Thomson Reuters. The court held that 2,243 Westlaw headnotes are protected by copyright. Headnotes are the short summaries of legal points that Westlaw editors write for court decisions.
It also held that using them to train ROSS’s search tool was not fair use, the US defense that allows some copying without permission. A contractor used the headnotes to write the memos that trained the tool. ROSS says the tool itself returned only passages from court opinions.
Three of the four factors in 17 U.S.C. § 107 weighed against ROSS:
- § 107(1), “the purpose and character of the use”: both companies used the headnotes to build a legal research platform. That made ROSS’s use “minimally transformative, at best.” The court added: “Unlike necessity, ease is not a justification for copying.”
- § 107(3), “the amount and substantiality of the portion used”: ROSS said it took only 0.08% of Thomson Reuters’s 28 million headnotes. The court said each headnote is a work of its own, copied whole.
- § 107(4), the effect on “the potential market for or value of the copyrighted work”: the court found “the market for licensing headnotes as text to train AI is rapidly developing.”
Only § 107(2), “the nature of the copyrighted work”, favored ROSS, because headnotes are largely factual.
What the ruling does not cover
“ROSS’s AI was not a generative AI, meaning it would not create any new expression,” the opinion says. It set the case apart from Bartz v. Anthropic and the OpenAI cases in New York, and called it “no more than an ordinary copyright case.”
It does not settle whether training a chatbot on books or news is fair use. Those fights continue in trial courts, like the music publishers’ suit where a judge said the publishers had cited no infringing Claude output. Europe runs the question through different law, as the EU court adviser’s test for chatbot answers showed on Monday.
What ROSS argues now
The petition says the panel judged ROSS’s business goal instead of its use of the headnotes. It also says no headnote reached the public and the markets were defined wrongly, AI Affairs reports. Law360 quotes the filing: “at every turn, the panel asked the wrong question.” We could not open the petition itself.
ROSS shut its platform in 2021, citing litigation costs, and has said it will seek Supreme Court review, LawSites reports. After the appeal, the case goes back to the Delaware trial court, where a tortious interference claim is still open.
What changes for a small team
If your team only uses chatbots like ChatGPT or Claude, little changes. It matters if you fine-tune a model or build a product on someone else’s content. Training on a rival’s material to build a substitute for it now has a US appeals ruling against it. The court also counted a training data licensing market against ROSS, though Thomson Reuters had licensed its headnotes to no one.
What to do this week
If you train or fine-tune anything, record where each dataset came from, under what terms, and whether a license was on offer. If you buy an AI tool built on legal, medical or other specialist content, ask the vendor in writing whether that data is licensed.
Sources
- Thomson Reuters v. ROSS Intelligence, No. 25-2153, Third Circuit opinion (GovInfo)
- 17 U.S.C. § 107, fair use (Cornell LII)
- AI Affairs: Ross Intelligence asks Third Circuit to rehear Westlaw copyright ruling
- Law360: Westlaw AI fight warrants full court review
- LawSites: 3rd Circuit unseals its opinion in Thomson Reuters case against ROSS
Get the weekly email
The week's most useful AI news, a tutorial and an interview.
Martin Scott
Martin Scott is the editor for business and policy at AI Observer, covering startups and deals, regulation, health and science, and AI agency profiles. He follows the money and the legal text, and separates what a rule says from what people claim it says. His style is analytical and plain-spoken.
Spotted a mistake? Tell us on the corrections page.
