3.2 KiB
Background
A one-page account of why the Inference Cooperative exists and how it got here. For how to join and use it, see Getting started; for the rules, see the Charter.
Mission
The Inference Cooperative exists to give people a real alternative for AI: a service owned and directed by the people who use it, rather than rented from an extractive industry. Members pay dues on a sliding scale and receive private AI chat and API access through infrastructure the members govern together.
Three commitments shape how we build it:
- Privacy. Our "private" models run in trusted execution environments (hardware-encrypted enclaves) with end-to-end encryption between your device and the model — the stack is designed so that even the co-op's own operators cannot read members' conversations.
- Sustainability. Members can choose models that run on 100% renewable energy, from providers with public environmental commitments. We self-host our own infrastructure, so every efficiency gain is ours to make.
- Self-governance. Decisions are made by members through policy governance, on our forum, using free, self-hosted tools wherever possible.
We are a small cooperative by design, fiscally sponsored by the nonprofit Metagov and funded through member dues on Open Collective. Membership starts at $5/month (suggested $15), which covers a $15 monthly usage allowance. The service costs what members put in, and no investor takes a cut.
How we got here
The cooperative launched in 2026, growing out of research on democratic digital infrastructure at the Media Economies Design Lab (MEDLab) at the University of Colorado Boulder. It began with a single privacy-first provider — Tinfoil, whose models run in encrypted enclaves — and a deliberately small toolset: a chat, an API gateway, and the basics for membership and payments.
From there it grew the way co-ops do: in response to what members asked for. A model picker grouped providers by what they stand for — private (encrypted enclaves) and green (renewable-powered: GreenPT and Regolo, the latter with a written zero-data-retention policy). A member dashboard made usage and spending transparent. A federated Matrix server, a forum, and a newsletter gave members places to talk to each other and to the project. The chat migrated to LibreChat, and the code that runs the member-facing services was opened up in public repositories.
In October 2026, members adopted the co-op's first charter — a democratic constitution setting out values, roles, and decision-making, with a commitment to redesign governance as the community grows past 250 members.
Where we're going
The near-term road: more model choices with strong privacy and environmental records, infrastructure hosted with values-aligned providers, and the working-group structure the charter describes. The longer arc: proving that people can govern their AI the way they govern a credit union or a land trust — and sharing what we learn with the co-ops that come after us.