As communities deploy AI systems they rely on but do not control, the need for sovereignty becomes urgent. We propose a framework for Open Governance of AI using Community Constitutions: rules written and amended by the people a system affects, and applied between an application and the AI model beneath it. Stakeholders can co-design and revise those rules in the open, without retraining the model or waiting for the AI provider to change it.

Four draft constitutions are published, free for anyone to fork and adapt. Regulators are increasingly asking those who deploy AI to show that their systems are accountable, and we are engaging with national regulators, legal partners, frontier model developers, and international governance agencies.