Wikimedia Foundation Proposes Middle Path in the Biggest U.S. AI Copyright Case

In a position that may seem surprising to many observers, the Wikimedia Foundation — the nonprofit organization behind Wikipedia — filed an amicus brief in the In re: OpenAI, Inc. Copyright Infringement Litigation case at the U.S. District Court for the Southern District of New York, and its message is direct: neither everything publishers say is right, nor everything OpenAI says is right.

The case, initiated in 2023 by The New York Times, alleges that OpenAI and Microsoft infringed copyrights by using protected content to train their artificial intelligence models. OpenAI and Microsoft argue that using publicly available material to develop AI systems falls under the concept of fair use, a legal doctrine that allows the use of protected material without authorization under certain circumstances, such as commentary, research, or transformation into something new.

Wikimedia wants judges to see the problem in two distinct parts. In the first stage — model training — the foundation argues that feeding copyrighted works into a large language model may indeed qualify as fair use. In the second stage — response generation — the organization draws a firmer line. A chatbot that reproduces protected material verbatim to users is a different animal, and Wikimedia doesn't want that behavior covered by the same legal protection.

The delicate balance

Wikimedia's position is fascinating precisely because it's uncomfortable. From the foundation's perspective, if the court rules that AI model training is illegal without a license, the practical effect would be to further concentrate AI development in the hands of large corporations — only companies with budgets for mass licensing could legally operate. For open-source developers, that would be devastating. Wikipedia, with its millions of volunteer editors, depends on AI-based tools to identify vandalism, detect copyright violations, and flag articles that need improvement. A total ban on AI training would ultimately strangle that ecosystem of tools.

But if the court rules that literal reproduction of content by chatbots is equally fair use, Wikimedia sees a different danger. A ruling that broadly favors verbatim reproduction would discourage people from sharing knowledge online. Wikipedia depends on the voluntary choice of people to publish freely. If any publicly published content can be absorbed and returned by a chatbot without limits, the incentive to contribute starts to look fragile.

> "The open web is a public good. We should not allow this behavior to become the 'new normal' for the people or organizations that maintain it," wrote Selena Deckelmann, chief product and technology officer at the foundation, after discovering that OpenAI agents had made unauthorized edits to Wikimedia projects in October 2026.

Licenses matter

One point Wikimedia insists on is that Wikipedia's content is free — the Creative Commons BY-SA 4.0 license allows use, adaptation, and sharing — but that doesn't mean there are no conditions. The license requires attribution to original creators and that adaptations remain under the same license.

The foundation argues that AI companies using Wikipedia content through retrieval-augmented generation (RAG) or similar methods should honor those license requirements rather than treating the use as universal fair use. Giving credit preserves the connection between the information people receive and the contributors who made it available.

Wikimedia also argues that companies collecting content for training or retrieval should respect Wikimedia's public API rate limits or use Wikimedia Enterprise services, which were specifically designed to manage commercial bulk access to the foundation's data. The idea is straightforward: if you want to use data at scale, pay for the proper service instead of overwhelming the foundation's servers.

A crowded courtroom

Wikimedia is far from being the only voice in this case. The Association of American Publishers filed briefs opposing broad fair-use protections for AI developers. The United States government also weighed in, filing a statement of interest in September 2026 supporting fair use for AI training.

Regardless of what the judges decide, Wikimedia's argument raises a broader question about how society wants the internet to function in the coming years. If human content can be absorbed by machines and returned without contributors having any control over that use, are we building a system where creative effort is simply a free input? But if AI training is entirely prohibited, we impoverish the tools that would make AI accessible to those who can't afford corporate licenses.

The question that remains is whether it's possible to build a truly open AI model without depending on a content ecosystem that, without adequate protection, disappears — or whether the only way to preserve the open internet is to accept that its maintenance has a cost only the largest can afford.

Sources: The Verge, Wikimedia Foundation, Cryptobriefing

✓ Independent sources cross-checked and verified before publishing