The Third Circuit unsealed its opinion in Thomson Reuters v. ROSS, affirming that ROSS infringed copyright by using Westlaw headnotes to train its legal research tool. The panel rejected ROSS's framing of the case as a referendum on the future of AI, describing it instead as an ordinary copyright dispute.
The court held that the headnotes cleared copyright's low originality bar and that ROSS's use was not fair use. It pointed to three key findings: ROSS built a direct commercial substitute for Westlaw, ROSS did not need the headnotes because it already had the underlying judicial opinions, and a market for licensing headnotes as AI training data was rapidly developing.
The decision is the first federal appeals court ruling on whether copyrighted material can be used to train an AI system, but the court limited its reach. Because ROSS's AI was not generative and only returned passages from existing opinions, the panel distinguished recent district court decisions in Bartz v. Anthropic and In re OpenAI that found training large language models can be transformative.
ROSS shut down its platform in 2021, citing litigation costs. The case began in 2020, when Thomson Reuters sued ROSS in Delaware federal court over the headnotes used to create training memos for the AI system.