Wikipedia editors have pushed back against plans from the Wikimedia Foundation to test AI-generated article summaries, powered by Aya, the open-weight AI model from Cohere. The non-profit has now paused the project. The decision came after a swift and overwhelmingly negative reaction from its community.
As first reported by 404Media, the plan involved a two-week, opt-in trial on the mobile version of Wikipedia. But the volunteer editors who build the encyclopedia met the idea with immediate and fierce opposition. The project's discussion page became a torrent of rejection. It included simple comments like "Yuck" and blunt declarations like "strongest possible oppose" and "Absolutely not."
One editor argued that a test would cause "immediate and irreversible harm to our readers and to our reputation as a decently trustworthy and serious source." They noted that Wikipedia has built its name on being sober and reliable, not flashy. Another feared it would destroy the site's collaborative model. They argued that while the "collective mass" of human editors "evens out into a beautiful corpus," the AI would install "one singular editor with known reliability and NPOV [neutral point-of-view] issues" at the very top of an article. That same editor also noted the following:
I feel like Simple Article Summaries (SAS) are contrary to a lot of things readers want in an encyclopedia. Readers come to the site trusting that we can give them all the information they want, while (crucially!) substantiating everything we say with sourcing and adhering to NPOV. While other readers could feel differently than I when I decided to join this community, without these two things, Wikipedia would be just another site.
For context, this is what AI-generated summaries on the platform was supposed to look like:
It is not hard to see why they are so protective. The editors' fears are grounded in recent and very public failures of AI features from tech giants. For example, Google's AI overviews recently hit 1.5 billion monthly users. The feature became a laughingstock for telling people to put glue on their pizza and that a dog had played in the NBA. This is the kind of humiliating error Wikipedia's community is desperate to avoid, as it would undermine two-plus decades of careful work. We also saw the potential for reputational damage back in January. That was when Apple's AI feature falsely generated a notification claiming that Luigi Mangione had died by suicide. The man was actually alive and in custody.
On the site's technical discussion page earlier today, Marshall Miller (MMiller), a Senior Director at the Wikimedia Foundation, posted an update acknowledging the feedback. He admitted, "It's clear we could have done a better job introducing this idea," and confirmed the experiment was paused. The Foundation says the goal was to explore accessibility for different readers. While this specific test is off the table, the organization still wants to use new technologies. Miller ended with a promise: "We do not have any plans for bringing a summary feature to the wikis without editor involvement."
A WMF spokesperson also told 404Media that though the feature has been paused, the foundation is still interested in AI-generated summaries. The spokesperson insisted the goal was to eventually build moderation systems where "humans remain central" and called this kind of backlash feedback part of what makes Wikipedia a "truly collaborative platform."
9 Comments - Add comment