Creating a Toolkit for Community Assemblies on AI

Community assemblies provide space for public dialogue on AI. There are many forms of community assembly. We are focussing on a model that generally involves:

• 🕐 4 - 16 hours of participation (Usually over 1 - 2 days)
• 👥 Intentional invitations: aiming for diversity, usually involving 15 - 50 people
• 📚 A structured agenda and learning materials
• 💬 One or more discussion modules, often with common questions
• 📝 Structured feedback activities to share individual and consensus views and actions

Over the coming months, the Citizens’ Track on AI aims to support community assemblies across the world that can feed into the next UN Global Dialogue on AI Governance, the 2027 Geneva AI Summit, and local and national AI governance processes.

To enable this, we’ve been running a co-design sprint over October 2026 to create a Community Assembly toolkit, and we ran a series of events in September to feed into the toolkit design.

The Co-design Sprint

We know that many other people are also thinking about the design of resources like this, and that effective resources need to be adaptable to different cultural, linguistic and local contexts. That’s why we’ve been working with a fantastic cohort of collaborators to join us over October who can bring experience and expertise related to:

• Building critical AI literacy amongst non-experts;
• Designing participation and deliberation activities;
• Current issues in AI governance;
• Facilitating community assemblies.

The sprint is taking the form of five 90 minute online meetings (weekly) and asynchronous working over the course of October, and we’ll be sharing results soon.

Workshop reports

We’ve run a series of online and in-person workshops in September and October in support of the co-design cohort.

Use the toolkit