A Delaware court has handed Thomson Reuters a landmark victory in what is being considered the first major artificial intelligence copyright case in the United States. The ruling found that legal AI startup Ross Intelligence infringed on Thomson Reuters' copyrights by using protected material from its Westlaw platform.
US District Court Judge Stephanos Bibas issued a summary judgment rejecting all of Ross Intelligence's defenses, particularly their claim of "fair use" - a doctrine that sometimes allows copyrighted works to be used without permission.
The case stemmed from a 2020 lawsuit where Thomson Reuters alleged that Ross Intelligence had reproduced materials from Westlaw to build a competing AI-powered legal research tool. The startup had used Westlaw's headnotes, which are copyrighted summaries of key legal points, to train their AI system after Thomson Reuters declined to license the content.
Judge Bibas emphasized that Ross Intelligence "meant to compete with Westlaw by developing a market substitute," which heavily influenced the ruling against fair use. The court found potential infringement of over 2,200 headnotes.
"We are pleased that the court concluded that Westlaw's editorial content created and maintained by our attorney editors is protected by copyright and cannot be used without our consent," said Thomson Reuters spokesperson Jeffrey McCoy.
The ruling has major implications for ongoing AI copyright disputes. Cornell University professor James Grimmelmann noted, "If this decision is followed elsewhere, it's really bad for the generative AI companies."
The financial toll of the litigation had already forced Ross Intelligence to shut down in 2021. A jury will now determine damages based on which of Thomson Reuters' copyrights remain active.
This verdict could impact dozens of similar lawsuits currently in the US court system involving major AI companies like OpenAI and Google, who are facing copyright claims from various content creators. The case sets a precedent for how courts may view the use of copyrighted materials in AI training.