Wikimedia Foundation files amicus brief in OpenAI copyright case, backing fair use for training but not RAG

The Wikimedia Foundation has filed a friend-of-the-court brief in the consolidated copyright litigation against OpenAI in the Southern District of New York, lodged on 30 October 2026. The case combines twelve lawsuits, the best known being The New York Times' action against Microsoft and OpenAI, which the paper began in 2023. The Foundation, a US 501(c)(3) nonprofit with offices in San Francisco, runs Wikipedia. It sides with neither party and says it is speaking up for open knowledge; its chief executive, Bernadette Meehan, says it wants to defend the free, human-made knowledge Wikipedia depends on.

“Any conversation around fair use and copyright in the age of AI cannot be limited to for-profit organizations: it must include stakeholders across the entire online information ecosystem,”

Stephen LaPorte, General Counsel, Wikimedia Foundation, in the Foundation's announcement · wikimediafoundation.org ↗

The two main sides are far apart. The Times says AI firms may not use its content for any purpose, training included, without permission. OpenAI and Microsoft reply that training and deploying models on publicly available data is fair use. Wikimedia splits the question in two. On training, it argues the use can be fair when the resulting model is something new and different from its sources, and need not meet the source's licensing terms. On retrieval-augmented generation, the technique that pulls in material to ground an answer, it says copying and returning content is not fair use, and companies should respect publishers' licences and give attribution under CC BY-SA. In short, it would let model training stand while holding firms to the terms of the content they return.

Wikimedia has its own stake in the answer. Wikipedia holds more than 67 million articles in about 300 languages, written by nearly 250,000 volunteer editors, and is among the highest-quality datasets used to train large language models. Its text is published under Creative Commons CC BY-SA 4.0, which asks users to credit the authors and share adaptations on the same terms. Its general counsel, Stephen LaPorte, said that Wikipedia is built through a collaborative process that needs a strong base of open, accessible human knowledge. The Foundation adds that scrapers should follow its public API rate limits or use its Enterprise offerings.

Its worry is about both extremes. A ruling that is too narrow on fair use could shut out open-source AI developers and limit the tools its volunteers use, while one that is too broad could weaken the open web by thinning out the secondary sources Wikipedia depends on. If training were ruled not to be fair use, the Foundation says, AI development would probably be left to the biggest and best-funded firms.

The Foundation has been before the courts on such questions already, having backed a Supreme Court challenge to Texas and Florida online-media laws. Engadget notes that summary judgment is expected soon, and that this is one of more than a dozen AI copyright cases in the courts, so the outcome could set precedent.