A world where the knowledge AI is built from stays legible to the people it came from, and the layer that coordinates it cannot be tilted by anyone running on it.
AI Commons builds and stewards that layer, beginning with a public record of what AI systems are built from and what they owe to.
The pieces already exist. Archives, libraries, universities, standards bodies, and open communities hold the material these systems depend on. What they don't have is a shared way to see how it moves, who used it, or what came of it. The work doesn't add up the way it should, because nothing records how it connects.
We build that record and hold it neutrally, which means we can't be in the business of selling position within it, and we can't be owned by anyone running on it.
AI Commons began in 2016 out of discussions among people and organisations working on AI for the common good. Since then we have worked across the ITU's AI for Good Summits, the G7 Hiroshima Process, the Global Partnership on AI, the World Economic Forum's Presidio Recommendations, and the Global Data Pledge.
For most of that time we seeded ideas and others carried them forward. The Registry is a different kind of work. A common language is only a language if the people who have to use it agreed to it. So, this one is being built in the open, with the institutions that hold this material and the communities whose knowledge is in it.
A coalition of governments, memory institutions, publishers and rightsholder collectives committing to build a consent-aware, provenance-rich commons, each at the level their holdings and authority allow.