Agenda

Much of the AI governance landscape is converging on a shared agenda: making AI systems safe, lawful, and accountable. These are goals regulation is built to pursue, through standards, requirements, and enforcement that apply across systems alike. Beyond them lies a second set of concerns that an approach through regulation does not reach as well, because the outcome is not a uniform threshold but a distribution that takes shape between parties: what a system consumes to run and whether the effects it produces are worth that draw, who captures the value AI generates and who bears its costs, whose data, labour, and compute a system is built on and on what terms, who has a real say in what it is designed to do, and whether the communities it acts on keep standing as its terms change. Such concerns cannot be settled by requirements set in advance for all AI systems within a jurisdiction; they are configured dynamically, as value chains form and relate to their social, economic, and natural environments, often cross-border. This is the domain of private ordering: the configuring is done through business decisions, taken privately and in commercial interest, with no party charged with the public-interest stakes. Each party answers for its own deal; no one answers for the pattern those deals add up to, and so the shape that forms across the economy appears as nobody's responsibility. Yet it is a governance domain in everything but vocabulary: decisions of real public consequence are taken in it continuously, with no shared instrumentarium for making their public consequences legible and governable.

The AI Commons (AIC) Framework is a proposal for those instruments. It draws on the Creative Commons and open-source precedent — standardised, voluntarily adopted terms that became systemic through accumulation and ordinary legal enforceability, without statutory mandate — and extends the same mechanism to the conditions of AI deployments. An adopter composes a commitment from a set of profiles — Reciprocity, Sustainability, Openness, Governance, Access, Value — and writes it into the contracts a capability already travels under. Once there, the commitment can extend as a requirement on others in the value chain, so a standardised commitment at one node can become a lateral requirement at the next, and coordination accumulates without any central mandate. Each profile sets the terms of one relation between a deployment and the world around it. A system is built on data, labour, and knowledge that others supplied; Reciprocity recognises those upstream contributions and returns part of the value the system generates to the people behind them, whose continued work it still draws on. It consumes energy and compute to run and leaves an environmental footprint; Sustainability measures that impact at each node and weighs it against the real-world effects of what the system does, holding the deployment to a net balance that is neutral at the least and regenerative where it can be. Openness requires operational transparency, so that an independent party can examine how the system actually behaves. Governance gives the communities and stakeholders a system materially affects a standing place in the decisions that shape how it works and how it changes. Access keeps the capability openly available, across user types and regardless of ability to pay. Value routes a share of the gains automation produces into an unconditional basic income that, aggregated across many adopters, reaches the wider community.

The framework's wider relevance to WSIS is that, because these are contract terms, they work inside the legal and commercial structures every country already has, and travel through cross-border value chains without requiring jurisdictions to align their laws. That makes the framework an implementation pathway that does not depend on national regulatory infrastructure: a procurer or an innovator in any region can begin to govern AI's impact now, without waiting for regulation to reach these questions. It is one concrete route toward the goal this week's gathering shares — an AI transition whose benefits reach all, not only the most technologically advanced.