Background
Thomson Reuters publishes Westlaw, the dominant legal research platform. A central feature of Westlaw is its database of “headnotes” — short, editorially written summaries of legal points from court decisions, organized under a proprietary key numbering system. Thomson Reuters employs lawyers to write these headnotes, and argues they are original copyrightable expression.
Ross Intelligence was a startup that built an AI-powered legal research tool intended to compete with Westlaw. Ross licensed Westlaw headnotes from a third-party data provider and used them as training data for its AI system. The goal was to enable Ross’s AI to answer legal questions in natural language by learning from the patterns in the Westlaw headnotes.
Thomson Reuters sued for copyright infringement. Ross raised fair use as its primary defense, arguing that training an AI on text is transformative — the output is a machine-learning system, not a reproduction of the original text.
The Court’s Holding
Judge Bibas granted partial summary judgment for Thomson Reuters, finding that Ross’s use of the headnotes was not fair use. The court analyzed all four statutory fair use factors (17 U.S.C. § 107) and found that on balance they weighed in Thomson Reuters’ favor.
On the critical first factor — purpose and character of the use, including transformativeness — the court found Ross’s use was not sufficiently transformative. Unlike cases where a secondary work adds new expression, commentary, or insight, Ross used the headnotes to build a product that directly competed with Westlaw’s core function: answering lawyers’ legal research questions. The use was commercial and the product served the same market as the original.
On the fourth factor — market effect — the court found Ross’s tool would substitute for Westlaw in the legal research market, undermining Thomson Reuters’ market for the headnotes and any potential licensing market for AI training. This factor, which the Supreme Court has called the most important, weighed heavily against fair use.
The ruling covered more than 2,000 headnotes that Ross had copied. The case is now on appeal to the Third Circuit, which heard oral argument on June 11, 2026 in the first federal appeals court review of AI training fair use.
Key Takeaways
- AI training is not automatically transformative. Courts will look at whether the AI output competes with the original work’s market. Using training data to build a directly competing product is a significant fair use red flag.
- Direct competition is decisive. Ross’s AI was designed to do what Westlaw does — answer legal questions. That direct substitution made it hard to argue the use was transformative or non-competing.
- Licensing markets for training data are legally significant. The court’s reference to potential licensing markets for AI training data signals that copyright owners can protect the right to license their content for AI training, not just for traditional uses.
- The Third Circuit appeal is the one to watch. As the first federal appellate review of AI training fair use, the Third Circuit’s decision in this case will set binding precedent across a three-state circuit and will likely influence every federal court grappling with AI copyright.
Why It Matters
Thomson Reuters v. Ross Intelligence is the most consequential AI copyright decision yet — and it found against the AI company. For the AI industry, which has largely operated under the assumption that training data use is fair use, this ruling is a serious challenge. It suggests that when AI training directly threatens the market for the copied work, the fair use defense is weak.
The decision is particularly significant because it came from a district court that had previously allowed fair use as a potential defense, then changed course after full briefing. It also precedes a wave of similar cases — involving Anthropic, OpenAI, and Meta, among others — that will be decided in the years ahead. The Third Circuit’s forthcoming ruling will be the next major milestone in what is shaping up to be one of the defining intellectual property battles of the decade.