Appeals court hands Thomson Reuters a landmark AI training win

A legal research company trained its AI on a competitor's database. That competitor sued. And this week, five years later, a federal appeals court said: no, that wasn't fair use — and it won't explain why just yet.

The case is Thomson Reuters v. Ross Intelligence, and the facts are almost quaint by today's AI standards. Back in 2020, Thomson Reuters accused Ross Intelligence of copying material from its Westlaw platform to train a competing legal search engine. Ross argued it was fair use — the same defense every AI company sued over training data has leaned on since. On Tuesday, a U.S. appeals court rejected that argument and upheld the ruling for Thomson Reuters. Reuters itself called it "a first-of-its-kind ruling by a U.S. appeals court in a wave of high-stakes copyright cases over AI training." The court's actual reasoning is sealed, so we don't yet know the legal logic — only the outcome.

Here's the part that should sting more than the ruling itself: Ross Intelligence isn't around to appeal, celebrate, or regroup. The company announced in December 2020 that it was shutting down its platform the following month, citing the cost of litigation. It tried to fight back with antitrust claims, accusing Thomson Reuters of illegally tying Westlaw's search tools to its case law database. A federal judge dismissed those claims in September 2024, ending that counterclaim. So by the time this appeals court ruling landed, the defendant had been gone for nearly five years. The win is almost entirely symbolic now — except that symbols are exactly what shape the next ruling.

And there are a lot of next rulings coming. PYMNTS reported in July that courts have split on where AI liability actually starts, and more cases are pending through 2026. This Thomson Reuters decision doesn't settle that split — but it's now a data point every judge, plaintiff, and AI company's lawyer will have to reckon with, sealed reasoning or not.

For authors, here's the stake. If training on copyrighted material without permission can be ruled against on appeal — not just at trial, but upheld — that strengthens the hand of anyone bringing a similar claim against the companies that trained large language models on books, including possibly yours. It doesn't guarantee a win. But it cracks open a door that AI companies have spent three years insisting was sealed shut.

What to do this week: if you're part of any pending author-AI litigation or considering joining one, flag this ruling to your counsel or your Guild rep now. It's the first appellate precedent of its kind, and precedent is the thing lawyers build arguments on. Don't wait for the sealed reasoning to leak — ask what it means for your case today.