A federal appeals court has rejected an AI company’s fair-use defense in a copyright case that has nothing directly to do with recorded music — and that is exactly why the music business is paying attention.
On September 29, the U.S. Court of Appeals for the Third Circuit affirmed a ruling for Thomson Reuters against ROSS Intelligence. ROSS had used thousands of Westlaw headnotes as part of the training process for an AI-powered legal research product. The court held that the copied material was copyrightable and that ROSS’s use was not fair.
What the court actually decided
The dispute involved editorial headnotes written by Thomson Reuters attorneys for Westlaw. ROSS used those headnotes to help create training memoranda for a competing legal-search system. In its opinion, the Third Circuit emphasized that ROSS was commercial, minimally transformative and building a substitute in the same market.
Reuters reported that the ruling is the first U.S. appellate decision to address fair use in an AI-training context. The full precedential opinion is also available through GovInfo and Justia.
Three factors worked strongly against ROSS: its use was commercial, it copied substantial protected expression, and the court found market harm because ROSS was trying to compete with the product whose editorial material it had used.
Why this does not automatically decide the Suno fight
The ruling is important, but it is not a shortcut to saying every AI-training lawsuit is now settled. The Third Circuit itself distinguished ROSS from generative-AI systems. ROSS’s product did not generate new expressive output in the way music-generation models do, and the court stressed the direct competitive relationship between ROSS and Westlaw.
Music Business Worldwide highlighted the decision because major labels are already litigating with AI music companies, including Suno. Those cases involve different technologies, different copyrighted works and different factual records.
So the decision is not a ruling against Suno, Udio or generative music. It is a new appellate data point showing that calling a use “AI training” does not automatically make the copying transformative or fair.
The market question may become the center of the music cases
For music rightsholders, one of the most consequential parts of the opinion is the court’s attention to market substitution. If an AI company uses copyrighted material to build a commercial tool that competes in a market connected to that material, the economic relationship can weigh heavily in the fair-use analysis.
Music lawsuits will ask different questions: whether training copies were necessary, whether outputs substitute for recordings, whether licensing markets for training exist, and whether a model creates sufficiently different expression. Courts may answer those questions differently across jurisdictions.
For independent artists, the issue is moving from theory to infrastructure
AI copyright arguments can feel remote until they touch an artist’s catalog. But the legal rules being formed now will influence whether future music models need licenses, what data can be used for training, how creators can opt out, and what remedies exist when protected work is copied at scale.
The lesson is not that one case has decided the future. It is that the courts are beginning to draw boundaries — and the music industry now has a real appellate decision to study rather than only predictions.
Featured image: Ales Nesetril / Unsplash. Illustrative editorial photography; the image does not depict the parties or court in this case.