Wikimedia backs AI training on copyrighted work, but draws a line at regurgitation

Wikimedia is urging courts to distinguish between training AI on copyrighted work and chatbots reproducing that work for users. Its argument in a brief filed in a group of AI copyright cases is that a sweeping ruling could harm either useful AI tools or the open web.

Watch Desk analysis

What happened

The Wikimedia Foundation, which operates Wikipedia, has taken a middle position in lawsuits involving publishers including The New York Times, OpenAI and Microsoft. It argues that copyright law should protect large language models training on copyrighted material, while not giving chatbots a free pass to regurgitate that material to users, The Verge reports.

The foundation says an overly broad ruling in either direction could have consequences: it could deprive Wikipedians of useful AI-based tools, or put the open web at risk. The report describes a brief filed on Friday and says Wikimedia is taking neither side in the lawsuits.

Why it matters

This is a useful distinction in a debate often presented as a choice between unrestricted AI training and shutting it down. Wikimedia’s position separates how a model learns from what a chatbot later gives back. That leaves courts with a harder question than simply deciding whether copyrighted material appears somewhere in an AI system.

The foundation’s stated concern also puts the stakes beyond the litigants. A ruling could affect how AI tools use material from the web, and whether services built on that material can return it to users.

Our read

Wikimedia’s argument is a reminder that “AI and copyright” is not one legal question with one tidy answer. The split between training and output gives courts a more precise line to consider, though where that line should sit is exactly what the cases will test. Even Wikipedia, the internet’s most dependable supplier of “citation needed”, has a stake in what comes next.

What to watch

• How courts address the distinction between model training and chatbot output.
• Whether the cases produce a ruling narrow enough to avoid the harms Wikimedia warns about.
• How the parties and courts respond to Wikimedia’s position.

Discussion spark: Should copyright law treat training a model on protected work differently from a chatbot reproducing that work for users?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.