Appeals court sides with Thomson Reuters in AI fair use case
It's the first US appellate ruling on fair use for AI training, and it lands in copyright holders' favor.
For two years, every AI company sued over training data has leaned on the same three words: fair use. This is the first time an appeals court has actually weighed in, and it didn't go the way Big AI wanted. Thomson Reuters won. That's not just one case closing — it's the first real precedent other courts will have to reckon with.
Every lawsuit authors have filed against an AI company — and there are a lot of them — has been waiting on one question: does training a model on copyrighted text count as fair use? Until now, that question had only been answered by district courts, with mixed results and no binding weight beyond their own courtroom. This week, an appeals court answered it for the first time, and the answer favors the copyright holder.
Thomson Reuters brought the case. It won. That single sentence is doing a lot of work here, because "first US appellate finding" is not a small phrase in legal terms — it means this ruling now sits above the district court level, and other courts weighing similar disputes will have to take it seriously, or explain why they're departing from it.
Here's what this means for you if you're an author with a claim pending against an AI company, or thinking about filing one: the legal ground under "fair use" just got firmer, and it firmed up in your direction. AI companies have spent two years telling courts that ingesting copyrighted books to train a model is no different from a human reading and learning from them. That argument just lost its first appellate test.
I want to be plain about what I don't know yet. I don't have the court's reasoning, the specific facts of the Thomson Reuters case, or how narrowly or broadly this ruling will be read by the judges handling the author class actions against OpenAI, Meta, and the rest. Those cases involve different facts — books versus legal databases is not a small distinction, and a court could draw a line between them. So don't treat this as a guaranteed win in your own case. Treat it as the best piece of ammunition your lawyers have had yet.
What to actually do this week: if you're part of one of the pending author class actions, ask your counsel — or the Guild if they're coordinating on your behalf — how this ruling specifically affects your case's fair use argument. Don't wait for a monthly update to land in your inbox; email now. If you're not yet part of a suit and you believe your work was used to train a model, this is the week to look into joining one, because the legal wind just shifted and plaintiffs' lawyers will be moving fast to cite this ruling in every active filing.
My take: courts don't often hand copyright holders a clean, precedent-setting win against an AI company. This is one. Don't let it sit as a headline you skimmed — make it a question you ask your lawyer this week.
Why it matters to you: This is the first appellate-level ruling on AI training and fair use, and it favors the copyright holder, giving authors' own infringement claims real precedent to stand on for the first time.
This story ran in the Coral Wire edition of Wed, Sep 30 2026, when the Coral Index stood at 58/100 (Tailwinds building).
Coverage — 2 sources
Every outlet found reporting this story.
Also in this edition
- Smashwords opens preorder sales to indie authorsWire — Fewer than 20% of Draft2Digital's 300,000+ authors currently use preorders at all, despite stronger launch sales
- NYT and publishers fight DOJ's support for OpenAI in courtLegal — The Justice Department backed OpenAI's side, and now the plaintiffs are asking the court to throw that support out.
- Amazon caps KDP authors at two new titles a weekPlatform — The cap lands the same week series writers are already pulling 96% of urban fantasy royalties while the rest of the market shrinks.
- Anthropic claim window doubles as OpenAI memos surfaceLegal — Filing deadline extends from thirty to sixty days as unsealed documents quote OpenAI on training GPT-3 with pirated works
- Little, Brown apologizes as Fourth Wing spinoff misses indie shelvesMarket — The publisher called itself "extremely sorry" after the new release failed to land at independent bookstores.
- Headline Home cuts two jobs in imprint restructureMarket — The redundancies mark a rare contraction inside Hachette's Headline stable.
- Utah's book ban list hits 44 titles, still growingLegal — Two districts, Davis and Washington, account for most of the state's 44 bans even as a lawsuit against the underlying law proceeds
- Kobo's 6.0 beta firmware wipes reader annotations and sideloadsWire — Kobo is telling affected readers not to roll back, warning the update makes missing data permanently unretrievable