# Wikimedia urges a middle ground in AI copyright cases

> Wikipedia’s operator says training AI on copyrighted work can be fair use, but AI systems that substitute for original work should need a license.

Oossa · 2026-10-30 · https://oossa.com/en/wikimedia-urges-a-middle-ground-in-ai-copyright-cases

The Wikimedia Foundation is asking courts to draw a line between training AI models on copyrighted material and using those models to reproduce material for users. In a legal brief filed in lawsuits brought by publishers including The New York Times against OpenAI and Microsoft, Wikipedia’s operator says the first use can qualify as fair use, while the second may require a license if it replaces readers’ visits to the original work.

Wikimedia says a ruling that treats all AI uses alike could harm both the open web and the nonprofit itself. Wikipedia is a major source of training data, the foundation says, but AI services can also draw people away from the sites and communities that create and maintain that knowledge. The organization further argues that services such as ChatGPT reproduce text covered by Wikipedia’s Creative Commons license without following its terms.

## Where Wikimedia draws the line

The foundation’s proposal turns on what an AI system does with the material. It argues that scraping websites and using copyrighted works to pre-train a model without a license should count as fair use when the model produces something new and different from its sources. But if a system returns copies of material in response to a user’s question and thereby substitutes for the original, Wikimedia says that use should require a license.

For Wikipedia content, the license terms include crediting the source and keeping adaptations under the same license. Wikimedia says that distinction matters because AI models also have practical uses for the encyclopedia: editors can use them to help identify vandalism, possible copyright violations and doxxing, or to spot articles that need improvement. The foundation argues that overly broad copyright limits could restrict such tools, while an overly permissive ruling could weaken the rights and licensing terms of human creators.

## A dispute with wider stakes

The New York Times alleges that OpenAI infringed its copyright by scraping its website for training data. OpenAI and Microsoft argue that training models is fair use. OpenAI has described the reproduction of copyrighted work as a rare bug; Microsoft has said expert analysis found Copilot rarely reproduces full sentences from news content used to ground its answers.

Wikimedia says neither side’s broad position fully accounts for the difference between training a model and delivering material that stands in for the original. Its brief does not decide how the courts should rule in every case, but asks them to assess those uses separately. The foundation says a careful approach could preserve both AI tools that support free knowledge and the licensing rights of people whose work those systems use.

## The facts

- The Wikimedia Foundation filed an amicus brief in lawsuits involving The New York Times, OpenAI and Microsoft.
- Wikimedia says using copyrighted work for model pre-training can be fair use when the model produces something new and different.
- The foundation says AI outputs that substitute for original material should require a license.
- Wikipedia’s Creative Commons terms require attribution and say adaptations must remain under the same license.

## Why it matters

For Wikipedia editors, the outcome could affect whether AI tools remain available to help flag vandalism, copyright problems and articles needing attention. Wikimedia’s position is that training and copying should be treated differently; how the courts will rule, and what that would mean for specific AI services, remains unsettled.

## Sources & references

1. [Wikipedia’s operator says a bad AI copyright ruling could threaten the open web](https://www.theverge.com/policy/1022880/wikipedia-openai-new-york-times-copyright) – The Verge, 2026-10-30

Last updated: 2026-10-30