Meta description: Thirty representatives from eighteen countries met in Rome to govern AI through cooperativism. What's real and what's missing.
Between October 13 and 15, 2026, thirty cooperative representatives from eighteen countries met in Rome. The goal was to move from being users of artificial intelligence to being the ones who design and govern it. An AI cooperative is a structure where members control the models because one vote per person sets the technical and ethical priorities.
A declaration of principles is a promise. Architecture decides what actually happens. Who controls the models. Who handles the data and the infrastructure. A manifesto can say anything. The machine executes what its design allows. That's why the uncomfortable question is more useful than any item in the communiqué. Can cooperatives move from declaration to real implementation?
What's confirmed is limited but tangible. There's a shared vision. A community of practice beginning to take shape. And an announced roadmap. What doesn't appear is the money or the commitments that bind anyone. Without a described funding method, intentions remain just that. Intentions. Without the first pound. Hard to move forward.
The Rochdale case helps measure the distance. In 1844 the Equitable Pioneers wrote their own rules before anyone could impose them. One vote per person. Fair prices. Transparency. They started with minimal capital and four products. They grew because every member contributed and stayed on as an owner. As explored in Stones Don't Lie, they solved a concrete problem, buying flour without chalk in it, and the movement was born from there.
Here's where the difference weighs heavy. Rochdale tested its rules every day at the Toad Lane store. The cooperative AI roadmap still hasn't revealed what that initial member contribution would be. Platform.coop has been pointing out the same gap for a while now. A community-owned AI needs to survive the dominance of the big platforms. The knot remains compute and data. Whoever doesn't control them depends on whoever rents them out.
Governing AI requires three elements at once. Technical capacity to train or adapt models. Capital to pay for compute. And power to make the rules stick. The dominant platforms own all three. A cooperative with a charter of principles currently has only one. This is more complicated than it looks. The cooperative tradition offers one-person-one-vote, a democratic governance model proven over centuries. It can be a credible alternative with different implications depending on the region.
Writing from Mexico, I watch these dynamics with hopeful caution. Cooperatives know the distance between principle and cash register. Scale forces federation, as credit unions have done with shared services. Something similar could apply to common compute infrastructure, though it remains a hypothesis. Not an announced plan.
What would make this roadmap concrete are precise answers. Who decides on the models and under what voting and open-audit rules. What common fund gets created, with what contribution and timeline. Who owns or rents the compute and how members' data gets handled. And what binding obligations each cooperative takes on, with clear consequences if they fail to meet them.
I don't have all the answers. This is an experiment. Rochdale shows that writing your own rules is a founding act, as long as someone puts up the first pound. Rome took the first step.
Will cooperatives manage to build that shared infrastructure before others finish writing the rules for them?
