Wikimedia urges courts to distinguish AI training from chatbot retrieval

The Wikimedia Foundation filed its brief on October 30, 2026, in support of neither side in the copyright litigation.[1][2] The combined litigation includes 12 lawsuits against OpenAI, among them The New York Times’ case against Microsoft and OpenAI.[1] The Times filed that lawsuit in 2023, alleging that the companies’ use of Times content, including for model training, violates copyright.[1] OpenAI and Microsoft contend that training and deploying models on publicly available data is technological development covered by fair use.[1] The judge had not ruled on the fair-use question or on accepting Wikimedia’s brief.[4]

The Foundation argues courts should assess fair use separately for each way copyrighted works are used during AI development.[2] It says training a general-purpose language model is likely fair use when the model is something new and different from its sources.[1] By contrast, Wikimedia describes RAG as finding and copying online works, then returning that content to users through a chatbot.[1] It argues that RAG and similar methods grounding chatbot answers in copied works are likely not fair use.[1] The Foundation says companies using copied works this way should comply with the relevant platform’s or publisher’s licensing requirements.[1]

For Wikipedia content, the Foundation says its Creative Commons CC BY-SA 4.0 license requires attribution and requires adaptations to remain under the same license.[1] It says uncredited use should be limited to model training, while RAG and similar uses should provide attribution under the license.[1] Wikimedia also says scraping its projects should follow public API rate limits or use its Enterprise API offerings.[1] The Foundation says non-human scraping and crawling can affect its ability to keep its sites online for human readers.[2] It hosts 13 free-knowledge platforms, including Wikipedia.[2]

The Foundation says the ruling will be influential as other courts interpret copyright law for new technological uses.[1] It warns that an overly restrictive fair-use ruling could inhibit innovation and limit tools that benefit open-source projects such as Wikipedia, while a blanket fair-use ruling could reduce the availability of secondary-source material and weaken online knowledge.[1] Wikimedia says Wikipedia has more than 67 million articles in 300 languages and nearly 250,000 volunteers contribute each month.[1] It describes Wikipedia as a dataset used to train language models and as an information source for chatbots, search engines and voice assistants.[1]