Wikimedia urges courts to allow AI training but block verbatim copying

In short: The Wikimedia Foundation filed a court brief asking judges to protect AI training on copyrighted text, while stopping chatbots from copying articles word for word.

What happened

The Wikimedia Foundation, the nonprofit that operates Wikipedia, filed an amicus brief on Friday in major copyright lawsuits involving OpenAI and Microsoft. An amicus brief is a document from a group that is not part of the case, but wants to share its view with the court.

The lawsuits include claims from publishers such as The New York Times. Wikimedia did not take either side. Instead, it argued for a middle position on “fair use,” the legal idea that some copying can be allowed without permission in certain situations.

Wikimedia said courts should treat AI training as fair use in many cases. AI training is when a large language model learns patterns from lots of text, a bit like a student reading many books to learn how writing works. But Wikimedia said chatbots should not “regurgitate” articles, meaning they should not spit out long, recognizable passages from copyrighted stories.

Wikimedia warned that a simple, all-or-nothing ruling could cause harm in either direction. A ruling that is too strict could block useful AI tools for Wikipedia editors. A ruling that is too loose could encourage AI companies to ignore licensing terms and could weaken how information flows between writers and readers.

Why it matters

Wikipedia and related sites are widely used as a source of text for training AI systems. At the same time, Wikimedia says AI answers can pull people away from visiting Wikipedia, which can reduce support for the volunteers who keep it updated. How courts draw the line between “learning from” text and “copying” it could shape what AI tools can do, what publishers can enforce, and what information people can easily access online.

Source: The Verge AI