The most consequential copyright idea of October 2026 did not arrive in a new lawsuit. It arrived in an old one being reused. On October 23, 2026, news outlets suing OpenAI and Microsoft highlighted a recent Third Circuit decision — Thomson Reuters v. ROSS Intelligence — as they pushed back against summary judgment motions in closely watched copyright litigation. Their argument is that ROSS shows training AI on copyrighted works is not automatically fair use. For founders and investors, the deeper point is different: the ruling treats a license to train AI as a market copyright owners are entitled to enter.
What the Third Circuit actually decided
ROSS Intelligence built a legal research search tool. To train it, ROSS copied the full text of 25,000 Westlaw headnotes — short summaries of points of law written by Thomson Reuters' human editors — after Thomson Reuters had rejected its attempt to license Westlaw content. ROSS then used those headnotes to write training memos, even though its finished tool returned only passages from judicial opinions, not the headnotes themselves.
The district court, in a decision by Judge Bibas, held that ROSS infringed 2,243 headnotes and that the copying was not fair use. On appeal, the Third Circuit affirmed in Thomson Reuters Enterprise Centre GmbH v. ROSS Intelligence Inc., No. 25-2153, in an opinion dated September 29, 2026, by Judge Montgomery-Reeves, which was unsealed in early October and discussed again in filings on October 23.
The court's reasoning is what other cases will cite. On the first fair use factor, the court found the use "highly commercial" and "minimally transformative, at best," because both companies were building legal research platforms. It rejected the analogy to search indexing, holding that ROSS copied the headnotes because that was easier, not because it was necessary. As the court put it: "Unlike necessity, ease is not a justification for copying."
The fourth factor — effect on the market for the original work — reaches furthest. The court held that ROSS harmed not only the existing market for Westlaw but also a developing market for licensing headnotes as training data. By copying them, the court said, ROSS usurped Thomson Reuters's opportunity to enter that derivative market. That is the first appellate statement that an AI training license is a protected market in its own right.
Why the news outlets are citing it now
News outlets suing OpenAI and Microsoft filed their opposition to the defendants' summary judgment motions on October 23, 2026, and asked the presiding judge for oral argument. Microsoft, opposing a separate motion by authors, argued that using written works to train AI models is fair use. Amicus briefs are due October 30, with replies due November 20.
Background: how fair use usually works — and why ROSS is different
Fair use is a four-factor test: purpose of the use, nature of the work, amount copied, and effect on the market for the original. Courts have often been sympathetic to intermediate copying that enables a new product, such as search indexing. ROSS limits that sympathy: copying necessary to build something new differs from copying that is merely convenient, and the court called the dispute "no more than an ordinary copyright case." If defendant and plaintiff compete in the same market, and the plaintiff could have sold a training license, copying looks less like transformation and more like substitution.
What this means for founders and investors
For AI founders, the lesson is to treat training data as a licensed input, not a free resource to be justified after the fact. Founders should be able to show, dataset by dataset, where material came from, under what license, and what was excluded. A company that cannot produce that record will struggle in diligence, partnerships, and litigation — and ROSS shows that "we didn't show the material to users" is not a complete defense. Clean, documented licenses for high-value content are becoming a competitive advantage, and data pipelines should be able to add licensed content, remove contested content, and document both decisions.
For investors, ROSS reframes AI diligence around data rights: provenance, licensing coverage, and exposure to markets the company competes in. It also boosts licensing infrastructure: rights clearinghouses, provenance tracking, and dataset auditing. The caution is scope: ROSS is one appellate decision on specific facts, and other courts may weigh the factors differently. Treat it as a credible signal that the licensing market is real, not as a final verdict on AI training.
Angeled.com connects founders with investors.
Patently-O, October 2026: Ease Is Not Necessity — Third Circuit Affirms No Fair Use in Thomson Reuters v. ROSS (No. 25-2153, September 29, 2026).
AIooze, October 23, 2026: Unsealed — the appeals court reasoning against ROSS Intelligence.
AIStockWire, October 23, 2026: OpenAI lawsuit tracker, on the October 23 publisher filings and briefing schedule.
Read Next
Open Weights Are Not Open Business Models: Mistral Large 4 and the Licensing Trap
Courts Are Writing AI's Rules Now: Three Rulings Founders Should Read
Delhi High Court Orders 24-Hour Takedown of AI Deepfake Videos Targeting Chief Minister
EU Court Adviser Says AI Training Is Legal — but Gemini's Answers May Still Infringe Copyright
AI Regulation After the Midterms: What Founders Should Do Before the January Bills Arrive
DOJ Launches Super Intelligence Task Force to Police AI Misuse



