The Wikimedia Foundation filed a neutral amicus brief on October 30, 2026, in the copyright litigation involving OpenAI’s artificial intelligence models. The foundation distinguishes pre-training, which it generally considers fair use, from retrieval-augmented generation, which reproduces and directly presents content and should therefore comply with licensing requirements. A ruling deemed too restrictive could impede technological tools that benefit Wikipedia’s volunteers, while a ruling deemed too permissive could encourage publishers to restrict access to their content. For artificial intelligence companies, Wikimedia’s position is favorable to pre-training but more restrictive for direct retrieval, potentially increasing licensing costs for products that provide sourced answers. For publishers, it offers greater leverage against services that reproduce their work, without delivering a blanket victory on the use of training data.
Source: Read the original article
