The global AI dialogue must build infrastructure, not just consensus

As the United Nations prepares for its first Global Dialogue on AI Governance this July, the world finds itself in an unusual position: there is growing agreement on what thechallenges are, but very little agreement on how to address them.

There is some consensus across governments, industry, philanthropy, academia and civil society: AI systems should be safe and trustworthy. Power should not be concentrated in the hands of a few actors. Governance should protect human rights, strengthen accountability and ensure that those most affected by technological change have a meaningful role in shaping it.

Yet consensus alone does not produce action. This is where the United Nations has a unique opportunity: as a convener, coordinator and facilitator capable of helping governments, researchers, civil society, philanthropy and industry build the connective tissue that effective AI governance requires.

The success of the AI Dialogue should therefore not be measured by the strength of its final declaration. It should be measured by whether it leaves behind some durable mechanisms for learning, coordination and collective action.

As the leader of a foundation focused on the intersection of technology and society, I have the privilege of observing how many governments, researchers, philanthropies, civil society organizations and technology companies are each grappling with the implications of AI.

Across sectors and geographies, much of the conversation remains focused on a familiar set of questions: How do we ensure AI is trustworthy? How do we evaluate increasingly capable models? How do we ensure that countries in the Global South have a meaningful role in shaping governance frameworks?

These are important questions, and they deserve sustained attention. But they can also distract from a more fundamental challenge: who has the capacity to answer them, and through what mechanisms?

Today, a small number of frontier technology companies control technical talent, computing resources, evaluation capabilities and the narratives around AI development. Governments and philanthropic funders make large-dollar bets based on information provided by the very actors they’re attempting to govern. This dynamic creates not just an information asymmetry, but a fundamental power imbalance that undermines effective governance.

Simultaneously, the actors who should be coordinating – national governments, philanthropic foundations, civil society organizations, research institutions – operate largely independently from one another. Governments pursue parallel strategies.

Researchers work across disconnected networks. Civil society organizations struggle to access technical expertise. Funders support promising initiatives that often lack pathways for coordination or scale.

We find ourselves in a predicament of inaction not due to a lack of effort, but a lack of connective tissue.

And for many countries, the problem is even more acute. Most governance conversations centre US and Western Europe perspectives, leaving countries with fewer resources systematically excluded from designing the frameworks that will shape their technological futures. Viewed through this lens, AI governance is not suffering from a shortage of ideas so much as a shortage of infrastructure.

If AI continues to evolve at extraordinary speed, governance systems must become more adaptive, collaborative and responsive.

That begins with creating durable peer-learning networks. Countries facing similar policy challenges should be able to share experiments, failures and lessons in real time, rather than waiting for annual convenings or lengthy reports. Governance should operate more like an active learning system than a static framework.

It also requires mechanisms for rapid knowledge exchange. When significant developments occur – whether a breakthrough capability, a safety incident or an innovative regulatory approach – decision-makers need trusted channels to understand implications quickly and coordinate responses where appropriate.

Equally important is moving from principles to practice. Around the world, governments, cities and institutions are already experimenting with different approaches to AI oversight. Those experiments should be documented, evaluated and shared openly so that governance becomes cumulative rather than fragmented. Learning should travel faster than technology itself.

A critical but often overlooked component is technical governance capacity. Discussions about policy interoperability are increasingly common. Discussions about technical interoperability are not. Yet effective governance depends on the ability to evaluate systems, test claims and understand the technical realities beneath policy frameworks. Without independent evaluation infrastructure and shared technical resources, even well-designed regulations can become difficult to implement.

Finally, governance infrastructure must create space for independent voices. Too often, the organizations building AI systems also shape the narratives surrounding them. A healthy governance ecosystem requires independent research, independent evaluation and meaningful participation from workers, educators, artists, Indigenous communities and others whose futures will be shaped by these technologies. Inclusion is not simply a matter of representation. It is a source of better decision-making.

No single institution can build this infrastructure alone.

Governments have a critical role in establishing rules, coordinating internationally and testing new approaches through implementation. Researchers can help translate complex technical developments into actionable insights. Civil society organizations can ensure governance reflects lived realities rather than abstract assumptions. Industry can contribute expertise and participate in standards development.

Philanthropy also has a unique role to play. Many of the institutions required for effective AI governance do not fit neatly into market incentives or government mandates. Independent evaluation capabilities, technical capacity-building programmes and cross-sector coordination mechanisms often emerge only when supported by long-term, mission-driven investment.

Philanthropy can help create the conditions under which governance infrastructure develops: funding experimentation, supporting independent institutions and connecting actors who rarely work together despite shared objectives.

The AI Dialogue represents a rare moment when the international community has recognized both the urgency of AI governance and the inadequacy of existing frameworks. If AI governance is ultimately an infrastructure challenge, then Geneva should be remembered not as the place where the world reached consensus, but as the place where it began building the capacity to act on it.