Copyright owners challenging AI companies now have a ruling they can point to, but the decision does not settle the larger fight over training data. A U.S. federal judge found that Ross Intelligence infringed Thomson Reuters’ intellectual property when it used Westlaw headnotes to train an AI legal research platform.
The case matters because it arrives while more than 39 copyright-related AI lawsuits are moving through U.S. courthouses. Still, the ruling is not a simple roadmap for every plaintiff suing an AI company.
What the court decided
The dispute centered on headnotes from Westlaw, Thomson Reuters’ legal research service. Headnotes are summaries of legal decisions, and Ross was accused of using them to train its AI.
Ross marketed its product as a legal research tool that could analyze documents and perform query-based searches across court filings. Thomson Reuters argued that Ross had built that competing product using protected Westlaw material.
In a summary judgment, Stephanos Bibas, the judge presiding over the case, rejected Ross’ core defense. Ross said its use of the headnotes was transformative because it repurposed them for a different function or market. Bibas did not find that argument persuasive.
The judge concluded that Ross was repackaging Westlaw headnotes in a way that directly replicated Westlaw’s legal research service. In his view, the platform did not add new meaning, purpose, or commentary to the protected material.
That finding weakened Ross’ claim that its use qualified as transformative. Bibas also pointed to Ross’ commercial motivations, because the company sought to profit from a product that competed directly with Westlaw and did so without significant “recontextualization” of the protected material.
Why fair use was not enough here
Fair use often turns on context. In this case, the court focused on whether Ross had done something meaningfully different with Thomson Reuters’ protected work.
The answer, at least at this stage, was no. The product’s role as a competing legal research system mattered. So did the court’s conclusion that the Westlaw material was not transformed into a substantially new purpose.
Shubha Ghosh, a Syracuse University professor who studies IP law, called the decision a “strong victory” for Thomson Reuters. He noted that the trial will continue, but that Thomson Reuters had won summary judgment at this stage of the litigation.
“The trial will proceed, [but] Thomson Reuters was awarded a summary judgment, a victory at this stage of the litigation,” Ghosh said. “The judge also affirmed that Ross wasn’t entitled to summary judgment on its defenses, such as fair use and merger. As a consequence, the case continues to trial with a strong victory for Thomson Reuters.”
For companies relying on copyrighted material as AI training data, the ruling is a warning against assuming that training alone automatically makes the use lawful. But it is also a case tied to a specific product, a specific market, and a specific kind of AI system.
The limits of the ruling
The decision has already drawn attention beyond the Ross case. At least one set of plaintiffs in another AI copyright case has asked a court to consider Bibas’ decision.
Even so, it is not yet clear whether other judges will follow the same reasoning. One reason is that Bibas made a distinction between “generative AI” and the AI system used by Ross.
Ross’ system did not generate new content in the way generative AI tools do. According to the source article, it merely returned judicial opinions that were already written. That distinction may matter in other cases.
Generative AI is at the center of copyright lawsuits against companies such as OpenAI and Midjourney. These systems are frequently trained on massive amounts of public web content and can generate speech, text, images, videos, music, and more.
Most companies developing generative AI argue that fair use doctrines protect the practice of scraping data and using it for training without compensation or credit. They say they are entitled to use publicly available content for training and that their models output transformative works.
Copyright holders see the issue differently. Some point to regurgitation, where generative AI creates content that closely resembles material used in training.
What it could mean for AI copyright lawsuits
Randy McCarthy, a U.S. patent attorney at the law firm Hall Estill, said Bibas’ focus on the “impacts upon the market for the original work” could become important for rights holders suing generative AI developers.
That market-impact question is central because many AI disputes are not only about copying. They are also about whether an AI product competes with the owner of the original work.
McCarthy also cautioned that the ruling is narrow and may be overturned on appeal. His view was that the decision should not be treated as the final word on AI training and copyright.
“One thing is clear, at least in this case: merely using copyrighted material as training data [for] an AI cannot be said to be fair use per se,” McCarthy told TechCrunch. “[But it’s] one battle in a larger war, and we’ll need to see more developments before we can extract from this the law pertaining to the use of copyrighted materials as AI training data.”
Mark Lezama, a litigation partner at Knobbe Martens focusing on patent disputes, saw a possible broader reach. He said the reasoning could extend to generative AI in its various forms.
“The court rejected a fair-use defense as a matter of law in part because Ross used [Thomson Reuters] headnotes to develop a competing legal research system,” he said. “Although the court hinted this might be different from a situation involving generative AI, it’s easy to see a news site arguing that copying its articles for training a generative AI is no d ifferent because the generative AI uses the copyrighted articles to compete with the news site for user attention.”
That is the narrow but important opening for publishers and copyright owners. The ruling may help them argue that AI training can harm the market for the original work, especially when the resulting tool competes for the same users or attention.
But the decision does not erase the differences between Ross’ legal research system and generative AI platforms. The broader legal picture will depend on how other courts handle those differences, whether Bibas’ opinion survives appeal, and how judges evaluate claims of transformation, competition, and market harm in future cases.