Full Breakdown
Third Circuit Rejects Fair-Use Defense in Thomson Reuters vs. Ross AI Legal-Research Case
By Drooid · · How we work
Core Ruling and Immediate Impact
The U.S. Court of Appeals for the Third Circuit affirmed a Delaware district court’s partial summary judgment for Thomson Reuters, holding that Ross Intelligence’s copying of Westlaw “headnotes” was not fair use. The panel, led by Judge Tamika Montgomery-Reeves, found Ross’s non-generative legal-search tool to be a market substitute for Westlaw rather than a transformative use. The sealed opinion, issued on a Tuesday, will bind district courts in the circuit once unsealed.
Background & Context
Thomson Reuters sued Ross in 2020, alleging the startup copied roughly 2,243 copyrighted headnote summaries and obtained about 25,000 “Bulk Memos” from Westlaw to train its AI-powered search engine. The district court granted partial summary judgment to Thomson Reuters on February 11, 2025, rejecting Ross’s fair-use defense. The RIAA and the National Music Publishers’ Association filed a joint brief on the same date, arguing that training AI models on copyrighted works cannot be fair use when the product competes with the originals.
Data & Statistics
- Headnotes copied: ~2,243
- Bulk Memos obtained: ~25,000
- Amicus brief filing date: February 11, 2025
Official Statements & Responses
A Thomson Reuters spokesperson said the company was “pleased with the decision,” noting that Westlaw’s editorial content is protected by copyright and that Ross’s copying “did not constitute fair use.” Ross’s attorneys did not respond to requests for comment.
Criticism & Opposition
Brandon Butler, Executive Director of the fair-use advocacy group Re:Create, warned that the ruling “hits a dead end” for AI developers seeking fair-use defenses. He argued that the court’s analysis failed to distinguish AI training from other intermediate-copying cases.
Why It Matters / Impact
The decision is the first appellate ruling on AI-related fair use, drawing a line between tools that reproduce existing content and generative systems that create new expression. Legal experts note that the case does not establish a blanket prohibition on AI training but underscores the importance of transformation and market effect in fair-use analyses.
Verbatim Quotes
- “WHILE THE FACTUAL CONTEXTS OF AI LEGAL RESEARCH TOOLS AND AI-GENERATED MUSIC MAY DIFFER, THE LOWER COURT’S HOLDING ESTABLISHES KEY PRECEDENT THAT TRAINING AN AI MODEL ON COPYRIGHTED WORKS TO COMPETE WITH AND SUBSTITUTE FOR THOSE COPYRIGHTED WORKS CAN NEVER BE FAIR USE.” — RIAA AND NMPA
What's Next
The Third Circuit ordered the parties to submit proposed redactions to the sealed opinion within 10 days. Separate AI copyright disputes remain unresolved, including the Suno case, where parties are expected to move for summary judgment on fair use. The outcome of these cases will further shape the legal landscape for AI training data.