The Wikimedia Foundation 10 filed an amicus brief on 30 with the U.S. District Court for the Southern District of New York in the consolidated case In re: OpenAI, which includes the copyright lawsuit brought by 12 against OpenAI (most notably, The New York Times v. Microsoft and OpenAI). The Foundation does not take sides but proposes a middle-ground approach: large language models trained on copyrighted content to produce "new things" that are "different" from the original works should qualify as fair use and need not comply with licensing terms; however, "grounding" techniques such as Retrieval-Augmented Generation (RAG), which retrieve original text and deliver it to users, do not constitute fair use and must adhere to the licensing terms of each platform.
Confidence: High (95) · Event date 2026-10-30
Reasoning
- This is one of the key cases adjudicating copyright in AI training, and its ruling will influence other courts. The Foundation worries about errors on both sides: a complete victory for copyright holders could gatekeep even innovative uses, restricting AI tools used by Wikipedia editors for anti-vandalism and infringement checks; a complete victory for AI companies would effectively condone ignoring licenses, leading to a more closed web.
- For Wikipedia, this means chatbots citing its content must attribute according to CC BY-SA 4.0 and maintain the same license; the Foundation also requires that scraping comply with rate limits on public APIs, or use the paid Enterprise API.
Usage
- The full text of the Foundation’s statement is available at wikimediafoundation.org; teams building RAG or search-based Q&A products can use it to check their attribution and licensing handling of Wikipedia content.
Risks
- An amicus brief is merely a third-party opinion and does not bind the court; the timing of the ruling is unknown.
- On the same day, OpenAI also published an article excerpting court filings from various parties supporting “training as fair use” (the original 403 unread); the complete legal arguments of both sides must be assessed by reviewing the court filings themselves.