Wikipedia's Operator Tells Court AI Training Should Stay Fair Use - but Chatbots That Replace the Original Should Pay

The Wikimedia Foundation has entered the most consequential AI copyright fight in court, filing an amicus brief on Friday in the lawsuits pitting The New York Times and other publishers against OpenAI and Microsoft. The nonprofit behind Wikipedia took neither side - and criticized both.

"Wikimedia is concerned that the parties' positions in this case push for overbroad and simplified applications of the fair use doctrine" around AI, the organization wrote in its filing. A clean win for publishers, it fears, would let copyright holders "gatekeep" even innovative and transformative uses of their work. A clean win for the AI companies would be "a green light to snub licensing terms and fragment the flow of knowledge between human authors and audiences."

Wikimedia's proposed line is drawn between training and substitution. Scraping websites should not itself be copyright infringement, the brief argues, and using copyrighted works to pre-train LLMs without a license should qualify as fair use when the models provide "something new and different from the original sources." But when a system pulls copies of material in response to user queries and delivers them to users "while supplanting readers' engagement with those original works," Wikimedia says that output is not transformative enough - and should require a license.

For Wikipedia specifically, the licensing point cuts both ways. Wikipedia text is freely licensed under Creative Commons terms that require attribution and that any remix or adaptation remain under the same license. The foundation argues that services like ChatGPT reproduce that licensed text without adhering to the license terms - meaning the very AI systems that benefited most from the open web have, in Wikimedia's view, been skirting its rules. The filing also notes that AI systems siphon visitors away from the sites that produce the knowledge in the first place, undermining the communities that maintain them.

At the same time, Wikimedia does not want a ruling that chokes off the models it finds useful. LLMs, the brief notes, help Wikipedia editors fight vandalism, detect copyright violations and doxxing, and find articles that need improvement.

"A ruling that does not apply fair use principles carefully could threaten the sustainability of free knowledge, and stifle the development of open source technologies, ultimately impacting access to information for everyone," general counsel Stephen LaPorte said in a statement. CEO Bernadette Meehan added that expanding access to knowledge through AI "must respect the people, contributors, and communities that make reliable knowledge possible."

The underlying case dates to December 2023, when the Times sued OpenAI and Microsoft over training and output practices. OpenAI has called regurgitation "a rare bug," and Microsoft told the court in September that expert analysis found Copilot rarely reproduces full sentences from news content. Wikimedia's brief reframes the stakes beyond the two parties: the entity that operates one of the most-cited data sources in AI training is warning that a black-and-white ruling, in either direction, could damage the free-knowledge ecosystem that both sides depend on.