{"id":7212,"date":"2026-10-05T10:50:48","date_gmt":"2026-10-05T01:50:48","guid":{"rendered":"https:\/\/en.ai-creators.tech\/media\/?p=7212"},"modified":"2026-10-05T10:50:48","modified_gmt":"2026-10-05T01:50:48","slug":"thomson-reuters-ross","status":"publish","type":"post","link":"https:\/\/en.ai-creators.tech\/media\/business\/thomson-reuters-ross\/","title":{"rendered":"Thomson Reuters v. ROSS: Appeals Court Rejects AI Fair Use"},"content":{"rendered":"

For the first time, a U.S. federal court of appeals has ruled on whether using copyrighted material to train an AI system is fair use. In Thomson Reuters v. ROSS Intelligence<\/em>, the answer was no.<\/p>\n

In an opinion dated September 29, 2026, and released in full the following day, the U.S. Court of Appeals for the Third Circuit held that ROSS’s use of Westlaw headnotes to train its AI legal research tool was not fair use. The panel affirmed the district court’s partial summary judgment for Thomson Reuters on both questions certified for appeal: the headnotes are original enough to be protected by copyright, and ROSS’s copying was not excused as fair use.<\/p>\n

The decision will be cited in AI copyright disputes for years. It is also narrower than many headlines suggest. ROSS’s system did not generate text; it retrieved passages from court opinions. The court itself described the dispute as “no more than an ordinary copyright case,” and it took care to distinguish the generative AI cases now shaping the field.<\/p>\n

\n

Key takeaways<\/strong><\/p>\n

The ruling does not hold that AI training is infringement, nor that it is fair use. It holds that copying a competitor’s protected editorial work to build a substitute product is not made fair by the fact that the copies were used to train an AI, particularly when the underlying public-domain material was freely available.<\/p>\n

The court also found harm to a developing market for licensing Westlaw headnotes as AI training data. That finding is specific to this case, but it shows how courts may treat training-data licensing as those markets mature.<\/p>\n

For readers following Japan, the reasoning has a close parallel in the proviso to Article 30-4 of the Japanese Copyright Act, which excludes uses that would unreasonably prejudice the interests of the copyright owner.<\/p>\n<\/div>\n

\n
\n

Table of Contents<\/p>\n<\/span><\/div>\n