A US appeals court upholds a ruling for Thomson Reuters in its copyright lawsuit against Ross Intelligence, rejecting Ross' fair use defense for AI training

Developers using scraped web data to train AI now face higher copyright risk for their output.

A U.S. federal appellate court has upheld a lower court's decision, ruling that the AI company Ross Intelligence infringed on Thomson Reuters' copyright. Ross Intelligence was found to have improperly trained its legal research platform using Westlaw's "headnotes," which are summaries of key points in judicial opinions. Thomson Reuters had sued Ross in 2020, alleging the company "surreptitiously" obtained the headnotes from a licensee to quickly develop a competing product without expending its own resources. Ross had argued its use was protected by fair use principles, asserting its AI search engine was a transformative use. However, the U.S. District Court Judge Stephanos Bibas ruled against Ross in February 2025, stating that Ross used the headnotes for the same purpose as Thomson Reuters and aimed to compete directly with it. The Third Circuit Court of Appeals' decision, currently temporarily sealed, has sided with Thomson Reuters, marking a significant appellate ruling on AI training data.

This ruling sets a precedent for how copyright law applies to AI training data, potentially impacting the way AI companies can develop new products. By rejecting Ross Intelligence's fair use defense, the court signals that using copyrighted material for the same purpose as the original copyright holder, even for AI training, may be deemed infringement. This could force AI developers to seek explicit licenses for data used in training, significantly increasing development costs and potentially slowing innovation in AI research tools. The decision also highlights the growing tension between AI advancement and existing intellectual property frameworks, as more companies may need to justify their data acquisition methods.

The decision is a victory for content creators and publishers who have expressed concerns about their intellectual property being used without permission to build competing AI services. Organizations like Disney, which filed friend-of-the-court briefs, argued that a ruling favoring AI companies would disincentivize investment in original content creation. For legal tech companies and other data-intensive industries, this ruling may necessitate a review of their data sourcing and AI development strategies. Future AI development may lean more heavily on licensed or publicly available datasets, or require new legal interpretations for transformative uses.