Wikimedia brief: AI training can be fair use, RAG answers should not
What this answers
- What did the Wikimedia Foundation tell the court about AI chatbots and fair use?
- In a friend-of-the-court brief filed Oct. 30, 2026, Wikimedia argued that training AI models on copyrighted content can be fair use, but that retrieval-augmented generation (RAG), where a chatbot fetches and reuses current web content in its answers, should not be treated as fair use.
- Which court case is the Wikimedia brief in?
- In re: OpenAI, Inc. Copyright Infringement Litigation in the Southern District of New York, which combines 12 lawsuits against OpenAI, including The New York Times Co. v. Microsoft and OpenAI, per Wikimedia.
The Wikimedia Foundation, which hosts Wikipedia, filed a friend-of-the-court (amicus) brief on Oct. 30 in the Southern District of New York consolidated case against OpenAI and Microsoft (12 lawsuits, including The New York Times’ suit), per its own post. It argues that using copyrighted works to train a model can be fair use when the model is something new and different from its sources, with no licensing terms required.
For retrieval-augmented generation (RAG), where a chatbot finds content on the web and delivers it back to users, and similar “grounding” techniques, the Foundation says use of copyrighted works should not count as fair use, and companies should follow each publisher’s licensing terms. For Wikipedia that license is CC BY-SA 4.0, which requires credit. Gizmodo reports the brief runs about 5,800 words and calls both OpenAI’s and the plaintiffs’ positions “overbroad”. The Foundation is an interested party; Gizmodo notes it has paid licensing partnerships with Amazon, Meta, Mistral, Perplexity and Microsoft.
Why it matters: Wikimedia, whose content sits behind many chatbot answers, is asking the court to treat retrieval-based answers differently from training, which is the line answer-engine citations would be judged on.
The record: OpenAIMicrosoftAI training & copyright litigation
Via Wikimedia Foundation ↗ · Gizmodo ↗
Posted to the wire October 30, 2026. Edited by Joe Balewski.