Wikimedia Urges Middle Ground In Openai Copyright Cases
In cases involving The New York Times, OpenAI and Microsoft, the Wikipedia operator argues that AI training can qualify as fair use while outputs that replace original works may require licenses.
The Wikimedia Foundation has urged a court considering AI copyright claims against OpenAI and Microsoft to avoid an all-or-nothing ruling on fair use. In a legal brief discussed by Engadget and The Verge, the nonprofit behind Wikipedia argues for protecting some uses of copyrighted material in AI training while requiring licenses when AI services deliver content that substitutes for the original work.
Table of Contents
A dispute over training and output
The New York Times is among the publishers suing OpenAI and Microsoft over their use of published content. According to Engadget, the Times challenges the companies’ ability to use its work without permission, while OpenAI and Microsoft argue that training and deploying AI models on published material falls under fair use.
Wikimedia is not asking the court to side entirely with either party. The Verge reports that its brief warns against interpretations that would let copyright holders block innovative uses of their work. But the foundation also cautions that broad protection for AI companies could allow them to disregard licensing terms and weaken the connection between creators and readers.
Where Wikimedia draws the line
Wikimedia argues that using copyrighted works to train a large language model can qualify as fair use when the resulting model offers something new and different from its sources. Engadget reports that the foundation also sees a potential fair-use role for retrieval-augmented generation, in which a model draws on outside knowledge sources when producing an answer.
The foundation draws a distinction when an AI service reproduces existing material for users instead of transforming it. According to The Verge, Wikimedia says a license should be required when an AI response supplies copies of a work and replaces readers’ engagement with the original. Engadget similarly reports that Wikimedia does not support retrieval-augmented generation that repeats articles without attribution.
For Wikipedia content, licensing has particular importance. The Verge reports that Wikimedia says some AI services reproduce text available under Creative Commons without following its terms, including crediting the source. The foundation argues that freely accessible content is not necessarily free of conditions on reuse.
Why the ruling matters to Wikipedia
Wikimedia says AI systems rely heavily on its sites for training data, while AI-generated answers can draw visitors away from the communities that maintain them, according to The Verge. At the same time, the foundation says language models can help Wikipedia editors identify vandalism, copyright violations and articles needing improvement.
That tension underlies its warning about both possible extremes: strict limits on fair use could constrain useful AI tools and open-source development, while unrestricted reuse could undermine the sources those tools depend on. Engadget reports that a summary judgment in the Times lawsuit is expected in the near future. The court has not resolved the fair-use question described in the reports.
Sources
This story was compiled by AI from the reports below. Read the originals for the full details.