Federal Appeals Court Affirms Thomson Reuters' Landmark AI Copyright Victory Over Ross Intelligence

A U.S. federal appeals court has definitively upheld a landmark copyright infringement victory for Thomson Reuters against the now-defunct legal artificial intelligence startup, Ross Intelligence. This pivotal ruling, issued on Tuesday, establishes a significant precedent as the first federal appeals case to reject a fair use argument for the unauthorized use of publisher content to train a commercial AI product. The decision by Judge Tamika Montgomery-Reeves, a member of the 3rd U.S. Circuit Court of Appeals in Philadelphia, underscored the creative nature of Thomson Reuters' materials and Ross Intelligence's direct intent to compete by leveraging them for a highly analogous purpose. The court's finding was based on "undisputed evidence" demonstrating the inherent creative spark within Thomson Reuters' copyrighted works.

Crucially, Judge Montgomery-Reeves highlighted that this case does not directly address the complexities surrounding large-language models (LLMs) or other generative AI tools that create novel content. Ross Intelligence's AI-powered legal search engine operated differently; it primarily functioned by retrieving and presenting existing passages from judicial opinions, rather than generating original text. This specific operational characteristic, the judge noted, may temper the broader implications of the ruling for other AI copyright litigation. She cautioned against overstating the case's impact on the future of AI legal technology, stating, "In truth, this is no more than an ordinary copyright case."

Thomson Reuters, a global information services company specializing in legal, tax, accounting, and news products, initiated its lawsuit against Ross Intelligence in 2020. This was a period well before the widespread public awareness and accessibility of generative AI chatbots like OpenAI's ChatGPT, which emerged later. The case is recognized as the first known AI copyright lawsuit filed in the United States and shares thematic similarities with numerous more recent legal actions brought by news publishers and content creators against prominent AI companies, including OpenAI and Anthropic.

At the heart of Thomson Reuters' claim was the assertion that Ross Intelligence had unlawfully copied thousands of "headnotes." These headnotes are concise summaries of the key legal issues and rulings within specific court cases, meticulously prepared and published by Thomson Reuters on its highly regarded legal research platform, Westlaw. The company argued that these headnotes were then instrumental in training Ross Intelligence's own competing AI-powered legal search engine, thereby infringing upon their intellectual property rights. In an earlier February 2025 ruling, Judge Stephanos Bibas had initially ruled in favor of Thomson Reuters, rejecting Ross Intelligence's defense that using copyrighted material for the training of a commercial AI model constituted legally permissible fair use. The appeals court's affirmation of this decision reinforces the principle that the unauthorized appropriation of creative works for the development and training of AI systems can indeed constitute copyright infringement, establishing a significant legal benchmark for ongoing and future disputes within the rapidly evolving landscape of artificial intelligence and intellectual property law.