Wikipedia’s nonprofit draws a line through the AI copyright fight
Wikimedia supports some model training while arguing that answers drawn from publishers must respect their licenses.
Wikipedia’s nonprofit is resisting a simple choice between protecting publishers and allowing AI development. In an October 30 court filing announcement, the Wikimedia Foundation said it backs fair use for training that creates something different from the source works, while opposing that protection for retrieval-based answers that reproduce publishers’ material. That is its argument in the OpenAI copyright litigation, not a court ruling.
Its concern runs in both directions. Restrictive training rules could impede useful open-source tools; blanket permission could weaken the news and other sources Wikipedia’s volunteers rely on. Wikimedia describes retrieval-augmented generation, or RAG, as looking up material to ground a chatbot’s response in current information. Its proposed dividing line concerns how material is used, rather than whether the product carries an AI label. For Wikipedia, the foundation identifies CC BY-SA 4.0 as the relevant license.
What open licensing permits
The Creative Commons BY-SA 4.0 summary helps explain why openly available material still has conditions. It permits copying, redistribution and adaptation, including commercial uses. In return, users must provide appropriate credit, link to the license and identify changes. Attribution must not suggest that the original creator endorses the reuse. Open licensing therefore supplies permissions with obligations attached; it does not make the work ownerless.
The ShareAlike condition requires adaptations to be distributed under that same license. It also prohibits additional legal or technological barriers that would stop others exercising its permissions. However, Creative Commons expressly preserves applicable copyright exceptions and public-domain material. That qualification is central to understanding this dispute: reading a license alone cannot establish whether a particular use instead qualifies for an exception. The summary also warns that privacy, publicity or other rights may still affect a reuse.
Why humans remain central
The foundation’s position fits its April 2025 AI strategy, which rejected replacing Wikipedia’s human editors with generated knowledge. Its proposed uses instead include easing routine moderation work, helping editors find information, translating material and supporting newcomers. The purpose is to leave volunteers more time for judgment, discussion and agreement. These are stated priorities, rather than evidence that every proposed tool has been delivered or that automated output is reliable.
That strategy also prioritizes human agency, transparent development and models whose code or weights are open. For readers, the useful connection is that Wikimedia sees a place for AI while retaining people’s responsibility for the encyclopedia. Its own strategy authors describe editors researching, debating and building consensus as the distinctive ingredient. An answer-producing machine and the community that maintains its underlying knowledge are different parts of that process.