품앗이 위키

Data Sovereignty in Practice

5 min read

Data Sovereignty in Practice

Data sovereignty is one of the stated aims of Solidarity Intelligence: “a place where the traces of our cognition, our relationships and our memory stay ours.” This note sets aside the abstract argument — covered in Fictitious Commodities and the Intelligence Proletariat — and describes what that actually looks like in the working systems this movement runs, in Daejeon, South Korea.

One axis of sovereignty per organisation

The movement is organized so that no single organisation, and no outside platform, holds every axis of a broader livelihood sovereignty. Instead, each of several co-operatives is responsible for one piece:

  • Poomasi Consumer Co-operative is the corporate root, running local food stores in Daejeon’s Jijok and Gwanjeo neighbourhoods. It is the site where the rest of this is tested against real, everyday operations rather than staying theoretical.
  • National Local Food Co-operative Network holds the food sovereignty axis.
  • Social Innovation Education Institute, a social co-operative, holds the energy sovereignty axis and runs automation AI for citizen renewable-energy work.
  • Community Currency Co-operative, through Hanbat Pay Inc., which it wholly funds and operates, holds the monetary sovereignty axis — running a citizen-led local currency.
  • Local Bridge Institute for the Social and Solidarity Economy provides intermediary support, connecting and mediating across the wider co-operative sector rather than holding an axis of its own.

Splitting responsibility this way is itself a data-sovereignty choice: none of these functions is outsourced wholesale to a single vendor or a single platform that would then hold the combined data — food purchases, energy usage, currency circulation, member relationships — of the whole membership in one place outside anyone’s control.

Poomai: infrastructure the co-operative runs itself

The clearest working example is Poomai (품AI), the AI system built around the Jijok and Gwanjeo stores. It is described internally not as a chatbot that merely answers questions, but as a system that “handles work” — it sends real text messages to regular customers and prints real price tags, described as “physical AI on the web.”

Its technical structure keeps the pieces that matter under the co-operative’s own control:

  • Front door: a home site served through GitHub Pages behind a Cloudflare CDN, with a chat widget that requires no login.
  • Processing: a GCP server that the co-operative operates, running Poomai’s own logic — not a black-box SaaS chatbot platform.
  • Data: Supabase, holding sales, product, and member data; ChromaDB, holding the system’s Q&A memory; and Netforce, the point-of-sale system — all systems the co-operative administers rather than data handed wholesale to a third party as the cost of getting a chat widget.
  • Language models: Gemini serves the no-login public chat, and Claude Code is wired directly to the core team. As of 14 April 2026, an internal Claude-based AI agent has been running in production wearing Poomai’s identity.

This mirrors the three-layer approach described under Domain AI: the commercial models (Gemini, Claude) are used as “the sword we use now,” while the data infrastructure, the store operations, and the accumulated Q&A memory are the “irreplaceable core” that stays with the co-operative rather than living inside someone else’s product.

Shared knowledge infrastructure

Beyond any single system, the movement maintains its own knowledge infrastructure rather than depending entirely on external platforms to hold its institutional memory:

  • A wiki (wiki.poomasi.org) recording the movement’s thinking, decisions, and working knowledge.
  • A knowledge graph (graph.poomasi.org) connecting concepts across that knowledge.
  • A digital-twin ontology (ontology.poomasi.org) modelling the co-operative’s own operations.

These are described candidly as mostly Korean-language working tools, run day to day by the co-operatives themselves — not polished external-facing products, but the actual place the movement’s knowledge lives.

Money that stays local

The monetary sovereignty axis works the same way: Hanbat Pay, the citizen-led local currency, and a companion climate currency called Gru, are built so that money keeps circulating inside the community rather than draining out through a payment processor that has no stake in the community it serves.

Education as a sovereignty practice

Data sovereignty and domain AI are of limited use if only a technical few can operate them. The movement treats training as part of the same practice: gathering materials, from concepts through to hands-on practice, “so that an organiser who has never written a line of code can build AI with their own hands,” through the AI Organisers’ Commons (activist.poomasi.org). Sovereignty, on this view, is not just about where data sits, but about whether the people whose data it is can actually operate the systems built on it.

What this is not

This is not a claim of total self-hosting or total independence from commercial infrastructure — GitHub Pages, Cloudflare, Supabase, Gemini, and Claude are all, in different ways, external services. The sovereignty being practised here is more specific: keeping ownership of the data, the logic, and the relationships that matter — customer relationships, member data, institutional knowledge, currency circulation — inside organisations the community itself governs, while still making pragmatic use of the best available outside tools for the parts that do not need to be owned.

Korean-language sources

그래프 뷰

백링크