This week in Geneva, national government ministers, industry representatives and international diplomats are meeting to discuss the future of AI governance. The first UN Global Dialogue on AI Governance is a step towards a more inclusive and multi-stakeholder approach, and an opportunity to respond to the fragmentation of existing initiatives.
This is the first time that all UN Member States have a seat at the AI table. The Global Dialogue has the opportunity to address the structural equity deficits in AI governance to date, where a small number of countries have shaped the agenda and upwards of 118 countries have been excluded. It also provides a space to go further and make sure the voices of people affected every day by impacts of AI are amplified and given the opportunity to shape agendas and action.
But there are challenges. We are facing a democratic crisis in AI governance: a disconnect between powerholders and people that is slowing the adoption of and adaptation to beneficial uses of AI. Just as AI governance demands more of the global system, multi-stakeholder governance is strained to breaking point. The succession of International AI Safety, Action and Impact summits held in the UK, South Korea, France and India reflects a rapidly broadening agenda that is still light on coherence, binding commitments and accountability mechanisms.
More positively, participatory democratic innovation continues to grow across the world: grassroots public participation initiatives on AI governance are providing ways for people to connect and empower themselves, and produce vital evidence to support better decision-making by powerholders.
At this critical point, the Global Dialogue and the upcoming Geneva AI Summit 2027 must learn from other experiences and embed citizen participation from the outset.
With The World Summit on the Information Society and the AI for Good Global Summit happening in parallel to the Dialogue, we are likely to hear serious and heartfelt words about AI for all, and of democratising AI through widening participation. The learning from the AI summits is that it takes more than a few words to deliver a democratic vision of AI.
To arrive at an equitable future for everyone, the Citizens’ Track on AI Governance, proposed alongside the Dialogue, will contribute to establishing mechanisms to involve the people that AI is meant to benefit in contributing to and evaluating how it is developed, deployed and governed.
Building the infrastructure for citizen voice
The white paper – A Citizens’ Track on AI governance: Alignment, agency and accountability, which grounds the work of the Track – brings together a vision for ongoing public participation at the heart of AI governance. It sets out the institutional infrastructure and interventions needed to move democratic AI governance from rhetoric to reality.
The ‘stack’ proposed in the white paper is not merely a system map. It shows the two-way connective tissue needed to link powerholders and people, and contributes to the infrastructure of a ‘governance floor’: a set of minimum standards that asserts that we should not be making AI governance decisions without hard-won human rights and the voices of those affected, and those who have most to gain from AI, in the room.
Source: A Citizens Track on AI Governance: Alignment, agency and accountability
As the diagram above shows, building the Citizens’ Track requires action at four interconnected levels.
Community assemblies
At the grassroots level, support is needed for a scaling up of public, inclusive, deliberative and focused conversations around AI. Projects like Let’s Talk AI – a literacy and awareness campaign that also uses co-produced cartoons to shape healthy conversations around AI – or the French government’s Cafe AI programme – which has sparked hundreds of self-organised open table discussions on AI across the country – demonstrate ways to support informed dialogue. However, more resources are needed to both sponsor and coordinate community assemblies that are plugged into the global debate and are able to build capacity for local action.
Coordination hubs
National, regional and thematic coordination can help make sure community assemblies on AI are not one-off but are connected to local levers of power, as well as to each other. The white paper explores the potential of transnational deliberation, drawing on learning from the Global Assembly on the Climate and Ecological Crisis, which has brought publics together across cultural and language barriers to create a deeply pluralist conversation and generate greater solidarity between people across the world. The work at the level of coordination hubs also shows that participatory practice is an essential part of building a democratically legitimate interoperability layer in AI governance.
Institutional interface
This layer acts as the bridge between distributed processes and the global AI governance system: keeping public participation informed about governance debates and making sure public input reaches governance fora in the right place at the right time. The exact nature of the institutional interface must be developed by participants in context, and may take the form of a standing secretariat or network of partners working together to connect and embed public participation in AI governance.
Docking points
Emerging sources of AI governance power stand at the top of the stack. They require clear structured connectors, or ‘docking points’, through which public participation can shape dialogue, decisions and actions. While the white paper focuses on multilateral and multi-stakeholder spaces, it argues also for the need to influence industry and national governments’ AI efforts, as these too must answer critical legitimacy questions for the way they shape AI futures.
PAVE-ing the way for the Citizens’ Track
The work proposed by the Citizens’ Track is vital as the UN Global Dialogue explores how to govern AI in the context of profound evidence gaps.
The co-chairs of the Independent International Scientific Panel on AI, created to parallel the UN Global Dialogue, have been clear: robust scientific evidence often lags behind the issues when it comes to AI. But this position is informed by habituated perceptions of what makes evidence scientific, rather than what is the most useful evidence for the decisions at hand.
We can and do understand how impacts of AI are unfolding right now: evidence from public participation provides the kinds of agile insights that good decision-making needs.
In part, the need to connect different traditions of evidence from different epistemic communities might be bridged by research: in opinion polls, people around the world are expressing strongly and consistently that AI is not developing in ways that align with their hopes and expectations. Not only do we have increasing evidence of rising public concern, but also of what people actually want from AI governance. Research from the Ada Lovelace Institute demonstrates this shifting public sentiment: people are increasingly looking for robust laws and regulations – with support for AI regulation rising from 62% in 2022/23 to 72% in 2024/25 – to improve their comfort with AI technologies.
However, what the Citizens’ Track proposes is to go beyond understanding public attitudes and establish the mechanisms to ensure that people can be part of solving governance challenges.
The new Public AI Voice and Engagement (PAVE) case book forms the evidence foundation of the Citizens’ Track. It presents over 80 participatory and deliberative processes focused on AI in the last five years. These have included over 200,000 participants from more than 95 countries, and involved over 280 different organisations from governments, industry, academia and civil society.
The case book shows where different communities have expressed their views on the impacts of AI, and where the gaps in insight and evidence remain. It builds directly on the Public Voices in AI (PVAI) programme and maps out where public participation in AI has been taking place around the world, offering a curated view of participatory processes that foreground inclusive methods and can offer inspiration and lessons learnt for new practice.
The case book also reminds us of the amount of work that still needs to be done. The PVAI programme’s review of research on public feelings towards, and experiences of, AI found that ‘methodologies and reporting mechanisms do not adequately reach, include or address the views of diverse communities. This means the views of some groups of people are not represented in the evidence. The lack of global majority research in the sample is striking […]’.
The examples in the PAVE case book start to show how to bring people into high-stakes, highly technical, discussions, even at the global level. And we know that inclusive models of participation have the potential to build deep public legitimacy, cut through geopolitical and corporate gridlock, increase resilience to the capture of policymaking by private interests, and ensure that the communities most impacted by systemic transformation have a direct hand in shaping guardrails and goals. The challenge – which we, the authors of this post, have tried to address in numerous participatory projects – is that powerholders often don’t see the opportunity that participatory approaches present.
A collaborative journey to a critical moment
The response that we, the authors of this post, propose to the unique occasion presented by the UN Global Dialogue in Geneva is also informed by our learning from the Participatory AI Research and Practice Symposium, and our experiences at the AI Impact Summit in Delhi. We heard practitioners and researchers share thoughtful, nuanced approaches to community voice – but often the reality is that there’s little resource to communicate findings, and both methods and messages struggle to cut through the noise.
That’s why we will have space in the AI for Good summit to present the case book, and share the recommendations and messages put forward by participants from processes around the world. And why we’re calling for political will and alignment from policymakers to support the Citizens’ Track, as a means to help them to make sense of public priorities and understand the strength of public mandates for AI governance.
Alignment across differences will be essential. Between the hype about AI exceptionalism and the complaints that policy can’t keep up, AI governance is fragmented and in need of alignment, agency and accountability. We have navigated these tensions for ourselves: collectively over 50 years representing evidence of public voice and advocating for greater engagement of people in AI, data and digital technology governance. Through our commitment to community, evidence and political change, we’ve collaborated on multiple evidence-building and advocacy initiatives in recent years.
While our different models for change mean we step forward with different emphases, we do align around common goals, seek out ways to build collective agency and hold ourselves accountable for equity and inclusion in amplifying the voices of the people who could benefit most or are most affected by AI technologies, their applications and uses. Seeking a commitment from leaders to embed public participation into the architecture of AI governance, we need to resist fragmentation and step forward with what unites us.
This is why we have come together again, alongside many wonderful colleagues across other organisations and continents and across numerous activities under the umbrella of the Citizens’ Track. With political will, funder investment and an institutional infrastructure that goes beyond listening, the UN Global Dialogue can connect to the wealth of community assemblies and help pave the way to participatory AI governance.