Wikimedia pauses AI summaries trial after backlash

Well, it's been a week for AI.

Just a day after Apple announced its own internal report found its own AI systems were lagging behind its cohorts (we're talking the likes of ChatGPT etc) Wikipedia has been forced to issue a mea culpa to its editors over its use of AI-generated summaries.

As a result, The Wikimedia Foundation (that's the nonprofit organisation that runs Wikipedia) has hit pause on the experimental project.

The AI experiment that sparked debate

It all began with the Foundation quietly planning a two-week trial of what it called “Simple Article Summaries” for the mobile version of the site.

According to 404Media, The idea was to use AI to generate short, easy-to-read summaries designed to help readers engage with complex topics. The trial summaries would be clearly labelled as “machine-generated” and flagged as “unverified.”

The AI behind the summaries was an open-weight Aya model developed by Cohere.

Now, it's this wasn’t some hidden replacement for human editing, the summaries appeared at the top of articles as expandable boxes, with human contributors able to review and flag issues. But despite the safeguards, many editors saw red.

Editors fear damage to trust and brand

The pushback was swift.

Editors from around the world argued that even a limited trial risked harming Wikipedia’s reputation as a reliable, sober source of information, an especially sensitive concern in an era of misinformation and declining public trust in media.

“Just because Google has rolled out its AI summaries doesn’t mean we need to one-up them,” one editor wrote.

Another noted that AI content could send the wrong signal about editorial standards: “It reinforces the belief that unsourced, charged content can be added, because this platforms it.”

A pause, not the end

Wikimedia has made clear this isn’t the end of AI exploration.

In a Wikipedia summary (no less) explaining the project, a spokesperson for the foundation noted the pause reflects its commitment to listening to its volunteer community.

Future experiments, they said, will ensure stronger collaboration with editors before any AI-driven tools go live.

“Bringing generative AI into the Wikipedia reading experience is a serious set of decisions, with important implications, and we intend to treat it as such,” a Wikimedia project manager noted.

Plans for any AI features will require editor moderation and involvement at every step.

Keep on top of the most important media, marketing, and agency news each day with the Mediaweek Morning Report – delivered for free every morning to your inbox.