Wikipedia's owner walks into OpenAI copyright fight, warns both sides could lose

The Wikimedia Foundation wants a word with the judge — but not a side. On October 30, 2026, the nonprofit behind Wikipedia filed an amicus brief in In re: OpenAI, Inc. Copyright Infringement Litigation, a filing described as backing neither party. Its pitch: whoever wins, the reasoning matters more than the verdict.

At the heart of the brief sits a line Wikimedia wants courts to draw. Pre-training — the foundational stage where a model soaks up patterns from mountains of text — typically qualifies as fair use, the foundation argues. Retrieval-augmented generation, or RAG, is different: those systems fetch specific content and serve it straight into answers, which Wikimedia says should trigger licensing requirements.

The filing lays out two ways this can go wrong. A ruling that's too restrictive could kneecap tools that help the volunteers who build Wikipedia; one that's too permissive could spook publishers into walling off their content. Stephen LaPorte, Wikimedia's general counsel, said bad rulings could threaten the sustainability of free knowledge, while CEO Bernadette Meehan framed the fight as defending free and open human knowledge as vital to Wikipedia.

There's a lot riding on the line for a site with more than 67 million articles across over 300 languages — one of the largest freely available bodies of human knowledge ever assembled. Wikimedia's brief asks the court to weigh training and retrieval as separate acts, not one blurry whole.

For AI developers, the split cuts both ways. Treating foundational training as fair use is friendly ground. But retrieval is exactly how modern AI products deliver fresh, sourced answers — and if courts adopt Wikimedia's framing, those features could arrive with license fees attached. Publishers get the mirror image: no blanket win on training data, but real leverage over products that reproduce their work verbatim.

Wikimedia isn't suing and isn't defending. It's telling the court that a broad stroke in either direction ripples well past OpenAI — straight into how the open web keeps its doors open.