The Wikimedia Foundation has filed an amicus brief in the ongoing copyright lawsuit between The New York Times, OpenAI, and Microsoft. Staking out a middle-ground position, the non-profit argues that training large language models on public web data should qualify as fair use, but that generating near-identical, regurgitated text output infringes on copyright and licensing terms.
Wikimedia emphasized that its platforms represent crucial training data across the AI industry, but uncredited content generation threatens the contributor communities sustaining free knowledge. Concurrently, the foundation noted that LLMs offer valuable internal utility for Wikipedia editors, including detecting vandalism, copyright violations, and doxxing.
The brief cautions against absolute rulings for either side. Wikimedia warns that a complete victory for publishers could let copyright holders gatekeep transformative innovation, whereas a total victory for AI platforms could erode open-source licensing terms and threaten the broader digital ecosystem.
Why it matters
A legal compromise allowing fair-use training but restricting output regurgitation could set a vital precedent for AI data sourcing.
Foundations and open-web platforms risk traffic loss if AI chatbots output their content directly without preserving creator attribution.
AI developers must prepare for stricter compliance around model outputs to avoid infringement on Creative Commons and open licenses.
Source: theverge.com



