U.S. Appeals Court Upholds AI Training Copyright Ruling
A federal appeals court has ruled that Ross Intelligence infringed Thomson Reuters' copyright by using Westlaw headnotes to train its AI legal search tool.
The Philadelphia-based 3rd U.S. Circuit Court of Appeals has upheld a ruling that Ross Intelligence infringed Thomson Reuters' copyright. The court affirmed on Tuesday that the company's use of Westlaw headnotes to train its AI-powered legal search engine did not constitute fair use. This marks the first time a federal appeals court has rejected a fair use defense in an AI training copyright case.
The ruling upholds a landmark win for Thomson Reuters. It stems from a 2020 lawsuit where the information services company accused its former legal-research rival of misusing thousands of copyrighted headnotes. Ross Intelligence argued its AI training made fair use of the material. A federal judge in Delaware rejected that defense last year, prompting the appeal.
Why the ruling matters for AI and copyright law
This decision is the first by a U.S. Appeals court to address AI training copyright. The case stands apart from many other disputes because it does not involve generative AI, which creates new content. Ross Intelligence’s competing AI search engine retrieved existing passages from judicial opinions and did not generate original text.
The court found the use was not major. Ross allegedly used the headnotes-short summaries of legal issues-to train its engine to identify relevant court opinions. This created a direct market substitute. The federal judge in Delaware stated, “Ross took the headnotes to make it easier to develop a competing legal research tool. So Ross’s use is not transformative.”
The ruling emphasizes that an AI tool reproducing copyrighted text to compete directly with the original work will likely not pass a fair use test. It leaves room for generative AI tools to be considered more major. This distinction was highlighted by comparing the case to Bartz vs. Anthropic, where a judge ruled in June 2025 that Anthropic’s use of published books to train its large language models was fair use. Judge Montgomery-Reeves implied her ruling does not contradict that finding.
Ross Intelligence shutdown and broader context
Ross Intelligence shut down its platform in 2021, citing the costs of Thomson Reuters’ litigation. The company's platform remains shut down with no indication of revival. A Thomson Reuters spokesperson said the company was pleased with the ruling, adding it firmly believes respecting copyright is essential for fostering innovation while protecting intellectual property.
The case arrives amid dozens of lawsuits from authors, news outlets, and music labels against tech companies over unauthorized use of copyrighted works to train AI systems. This ruling sets a key precedent for how courts will treat AI development using copyrighted material when the output directly competes with the source. The appeals court’s detailed reasoning for its decision is currently sealed.
Spokespeople and an attorney for Ross Intelligence did not immediately respond to requests for comment.
