Wikipedia's Volunteers Threaten to Strike Over Layoffs

Wikipedia runs on the unpaid labour of volunteers, and this week a good number of them said they had had enough. After the Wikimedia Foundation dissolved the small engineering team that builds the features its community asks for, hundreds of editors began organising what could become the first coordinated strike in the encyclopedia's history.

What the Foundation cut

The Foundation disbanded its Community Tech team, the group responsible for implementing volunteer feature requests through a long-running programme known as the Community Wishlist. Five software developers and a manager lost their jobs. A week earlier, the Foundation had also let go of Brooke Vibber, the lead developer who had stewarded the MediaWiki software that powers Wikipedia since 2003, according to heise online.

Leadership defended the move by arguing that funnelling every community request through one dedicated team had stopped working. The breadth of the software it supports, they said, made that model impractical.

The revolt

Volunteers did not take it quietly. Within a week roughly 600 editors, including around 50 administrators, had signed a solidarity petition, one of the most-signed in the platform's history. By the end of May, The Register put the count past 800. A newly formed group calling itself Wiki Workers United has been gathering formal support. The proposed actions go well beyond a petition: an editing strike, a pause on the volunteer vandalism cleanup that keeps the site readable, and, most pointedly, technical measures to block or replace the Foundation's donation banners.

Where AI comes in

The anger is sharpened by money, and AI is a big part of where that money now comes from. The Foundation earns millions licensing Wikipedia's content to AI companies to train their models, through a commercial arm called Wikimedia Enterprise, and it sits on reserves approaching $300 million. To the volunteers who write and maintain the articles, the math looks galling. Their work is being sold to train the very systems reshaping the web, while the team that served their requests gets cut.

There is a deeper tension here that goes beyond one organisation's budget. Large language models are trained on Wikipedia more heavily than on almost any other single source. If the people who tend that source feel taken for granted and step back, the erosion would not show up in any model's benchmark scores for a long while. It would just quietly degrade the commons that so much of modern AI quietly depends on.