News · AI

A US appeals court rules that training a legal AI on Westlaw's headnotes was not fair use

The Third Circuit upheld the ruling against ROSS Intelligence. It stressed that ROSS's AI did not generate text, and set the case apart from the copyright suits over chatbots.

Catching up: the original came out on . This story was published later, when the site opened.

Grade 1 · Primary

Source: US Court of Appeals for the Third Circuit

What changed
On 29 September the Third Circuit affirmed that ROSS Intelligence infringed Thomson Reuters's copyright by using 2,243 Westlaw headnotes to train its AI legal-research tool.
Who it affects
Companies that train AI on copyrighted material to build a product competing with the material's owner, in the Third Circuit (Delaware, New Jersey, Pennsylvania).
What it does not tell you
What this means for chatbots. The court said ROSS's AI did not generate new text, and that the questions raised in generative AI cases did not apply.

The full story Check the source

A US federal appeals court ruled on 29 September that ROSS Intelligence's use of Westlaw's headnotes to train an AI legal-research tool was not fair use [1].

The Court of Appeals for the Third Circuit affirmed a District of Delaware ruling for Thomson Reuters, which owns Westlaw. It held that 2,243 headnotes were original enough for copyright protection [1].

"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" [1].

What it covers

ROSS built a search engine that answered legal questions with passages from judicial opinions. A contractor wrote about 25,000 training memos for it, using Westlaw headnotes to frame the questions [1].

The court found ROSS's use "minimally transformative, at best," because ROSS and Westlaw used the headnotes for the same purpose: a legal-research platform [1].

It also found harm in "the market for licensing headnotes as text to train AI," which it called "rapidly developing." ROSS took Thomson Reuters's chance to license them, the court said [1].

Three of the four fair-use factors weighed against ROSS; the second, the nature of the work, weighed slightly in its favour [1].

What it does not tell you

Whether training chatbots is fair use. "ROSS's AI was not a generative AI," the court said. It distinguished the claims made for systems that "generate original responses" in other cases, including the pending suits against OpenAI [1].

How the case ends. This was an early appeal on two questions the lower court sent up: originality and fair use. The rest of the case continues there [1].

Whether it binds other courts. It is the law in the Third Circuit only [1].

Sources and their grades

  1. US Court of Appeals for the Third Circuit, Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., No. 25-2153, Opinion of the Court. Read 5 October 2026. Grade 1 · Primary Filed 29 September 2026; 32 pages, read in full for the holdings. An interlocutory appeal: the case returns to the District of Delaware.

A grade says how much weight a source will take, from 1 (a body that can compel an answer and has nothing to sell) to 5 (somebody describing what they sell, or themselves). The standard sets out what each grade may be used for.

Filed under AI, Copyright

Comments

Commenting needs an account, made with nothing but an email address. A first comment is read before it appears.

More news

See all