Skip to content
Policy · Oct 30, 2026

Wikimedia Foundation urges courts to protect AI training while limiting model output reuse

Wikimedia argues fair use should shield LLM training on copyrighted works but require licensing when AI reproduces copyrighted material in response to user queries.

Trust74
HypeLow hype

1 source · cross-referenced

ShareXLinkedInEmail
TL;DR
  • Wikimedia Foundation filed an amicus brief in ongoing AI copyright cases arguing that training LLMs on copyrighted material should be fair use.
  • The nonprofit proposes that AI companies should license content when models output verbatim or near-verbatim copies of copyrighted works in response to user queries.
  • Wikimedia warns a rigid ruling could either block beneficial AI tools for editors or enable unlicensed reuse that undermines open knowledge ecosystems.
  • The filing distinguishes between transformative model training and output that supplants engagement with original works.

The Wikimedia Foundation, the nonprofit that operates Wikipedia, filed an amicus brief in ongoing copyright cases involving The New York Times, OpenAI, and Microsoft, arguing that courts should protect large language model training on copyrighted works under fair use. At the same time, the foundation contends that AI companies should be required to license content when models output verbatim or near-verbatim copies of copyrighted works in response to user queries. The brief takes neither side in the underlying lawsuits, instead staking out a middle ground aimed at preserving both open knowledge and fair compensation for creators.

Wikimedia’s filing warns that an overly rigid ruling could either deprive Wikipedia editors of useful AI-based tools or enable unlicensed reuse that fragments the flow of knowledge. The foundation highlights that Wikimedia websites have become critical sources of data for AI and machine learning training, but also notes that AI systems like ChatGPT reproduce text licensed under Creative Commons without adhering to licensing terms. Wikimedia argues that while LLM training should be considered fair use, output that supplants reader engagement with original works should require licensing to respect the terms of those licenses.

The foundation proposes a compromise: scraping websites should not constitute copyright infringement, and using copyrighted works to pre-train LLMs without a license should be fair use when the resulting models provide something new and different from the original sources. However, if models deliver copies of material in direct response to user queries in a way that substitutes for the original work, that should require a license. Wikimedia emphasizes that this distinction preserves transformative uses of copyrighted material while preventing AI output from undermining the communities and contributors that sustain open knowledge projects.

In a statement, Wikimedia general counsel Stephen LaPorte said a ruling that does not carefully apply fair use principles could threaten the sustainability of free knowledge and stifle open source technologies, ultimately impacting access to information. Wikimedia CEO Bernadette Meehan added that expanding access to knowledge through AI must respect the contributors and communities that make reliable knowledge possible, calling for a balance that benefits the entire information ecosystem.

Sources
  1. 01The Verge — AI — Wikipedia’s operator says a bad AI copyright ruling could threaten the open web
Also on Policy

Stories may contain errors. Dispatch is assembled with AI assistance and curated by human editors; despite the trust-score filter, mistakes happen. We correct publicly — every article links to its revision history. Nothing here is financial, legal, or medical advice. Verify before relying on any claim.

© 2026 Dispatch. No ads. No sponsorships. No paid placement. Reader-supported via Ko-fi.

Built by a person who cares about honest AI news.