Thomson Reuters Wins Possible Landmark Copyright Case Against AI Company
When it comes to raising expectations, few products have been presented with as much relentless hype as generative artificial intelligence (AI). Those touting AI have promised it will do everything from driving our cars for us to writing amazing movie scripts for Hollywood in seconds.
For the most part, AI results haven't yet provided the lofty returns some prognosticators had hoped. While there's been success in some fields, how AI gathers and uses information has come under fire. Questions about the legality of its methods have found their way into courts across the nation.
A possible landmark decision occurred on February 11, 2025, when a U.S. district court held in favor of Thomson Reuters' motion for summary judgment regarding claims of copyright infringement. The decision may act as precedent for the hundreds of other cases involving AI currently in the legal system.
How Artificial Intelligence Works
AI is based on programming computer systems to both spot and learn from patterns in data sets. Using algorithms and models, the systems rely on machine learning to accomplish human tasks such as speech recognition, generating images based on prompts, or making decisions.
AI development follows five basic steps to reach its goal. The process starts with input as data is gathered and sorted. This is followed by processing, as the system uses the algorithms and patterns it has been programmed with. In the outcomes step, it analyses the data to see if it matches previous patterns.
If the data doesn't match, the AI "learns" from it during the adjustments phase. Finally, the AI assesses what it has learned and provides predictions. Generative AI, such as ChatGPT or OpenAI, takes these algorithms and uses large language models to produce content instead of analyzing it. Whether this content is truly original lies at the crux of many lawsuits facing AI companies.
The method AI uses to gather data is called scraping and involves extracting information from websites. AI scraping often collects intellectual property as well, which raised a hue and cry from creators over copyright infringement. While many cases are pending, this first major decision on AI copyright issues may serve as a guiding beacon in exploring previously uncharted legal waters.
Off With Their Headnotes
Ross Intelligence, a legal AI startup, approached Thomson Reuters (TR). They wanted to buy access to the online data of TR's subscription-based Westlaw legal research platform for the training of their legal AI system. TR, realizing they'd be enhancing a future competitor, refused.
Ross Intelligence then went to a company called LegalEase that had paid access to Westlaw and brokered an agreement to buy "Bulk Memos." These contained legal questions with good and bad answers that included Westlaw's headnotes, which were used as data for the training of Ross Intelligence's AI.
TR filed a suit against Ross Intelligence alleging copyright infringement in May 2020. In his ruling in February 2025, Third Circuit U.S. Court of Appeals Judge Stephanos Bibas, who is overseeing the case, admitted he made an error in the decision he rendered in 2023.
In the 2025 ruling, Bibas issued a partial summary judgment that held Ross Intelligence had used the copyrighted material to position itself as a direct competitor of TR.
Bibas rejected all of Ross Intelligence's defenses citing fair use, saying that none of their "possible defenses hold water." Their use of over 2,200 of TR's headnotes indicated that Ross Intelligence knew that the data they were giving their AI to ingest was under copyright protection.
The reason the summary judgment was partial is because Judge Bibas held there could still be a factual dispute over which headnotes were registered copyrights or if some were in the public domain.
The Legal Precedent for AI or a One-Shot Ruling?
The decision is seismic for generative AI companies. It could portend a legal bulwark that can influence AI copyright violation lawsuits going forward. It may allow individual creators, such as artists who have had their work used by AI, to initiate legal actions where they might not have had the confidence to do so before.
One company that no longer has to worry about it is Ross Intelligence. Citing the cost of litigation, the startup ceased operation in 2021.
Related Resources
- Is Grammarly Generative AI? (FindLaw's Practice of Law)
- Bumpy Road Ahead for All in Adoption of AI Into the Legal Industry (FindLaw's Practice of Law)
- Intellectual Property Law (FindLaw's Learn About the Law)
