Ross Intelligence was a startup that built an AI-driven legal research tool meant to compete with Westlaw, the dominant legal database owned by Thomson Reuters. To train its system, Ross obtained “bulk memos” derived from Westlaw’s headnotes - the short editorial summaries that West attorney-editors write to capture the key points of law in a court opinion. Thomson Reuters sued for copyright infringement in 2020. Although the text of judicial opinions is in the public domain, Westlaw’s headnotes and its Key Number System taxonomy are editorial creations.
On February 11, 2025, US Circuit Judge Stephanos Bibas (sitting by designation in the District of Delaware) issued a memorandum opinion that revised his own 2023 decision. He opened with unusual candor: “A smart man knows when he is right; a wise man knows when he is wrong. Wisdom does not always find me, so I try to embrace it when it does - even if it comes late, as it did here.” Bibas granted most of Thomson Reuters’s motion for summary judgment on direct copyright infringement, holding that more than 2,200 of the headnotes were original enough to be protected and were copied. He then rejected Ross’s fair-use defense, finding the use was commercial and not transformative, and that Ross’s product was meant to compete directly with Westlaw - harming the market for the original work.
This was the first US court ruling to squarely address whether using copyrighted material to train an AI system can be copyright infringement, and the first to reject a fair-use defense in that context. Importantly, Ross built a search-and-retrieval tool, not a generative model, which the judge noted distinguished it from the wave of generative-AI cases. Bibas later paused the case to allow an interlocutory appeal to the Third Circuit.
Why business readers should care: the ruling signaled that training data is not automatically free to use just because the end product is “AI.” The decision turned on the editorial originality of the headnotes and on market competition - the same factors that drive the larger copyright fights over generative models.