Despite bringing together 180 policymakers, civil society advocates, industry leaders, and academics, the global AI governance process has created platforms for conversation but failed to translate participation into genuine agency for the communities most affected by AI systems. That's the central finding from a major multistakeholder convening held in Geneva on July 8, 2026, which revealed deep structural asymmetries in how AI policy gets made.

What Happened at the Geneva AI Governance Summit?

On July 8, 2026, the Global Network Initiative (GNI), the UN Human Rights Office (OHCHR), Business for Social Responsibility (BSR), and a coalition of civil society and academic partners convened "Governing AI Together," a full-day event at the Palais des Nations in Geneva. The gathering brought together 180 participants representing a cross-section of stakeholders: 70 civil society representatives, 39 from standards development organizations and foundations, 24 academics, 19 government representatives, 17 from industry, and 11 from think tanks.

Across six sessions, participants launched new initiatives on human rights due diligence in AI product design, held discussions with UN Global Dialogue on AI Governance (UNGDAI) co-chairs, debated how to operationalize human rights protections in AI systems, examined the power dynamics between Big Tech companies and AI startups in Asia, previewed an upcoming Global South AI Safety Report, and co-designed a "Citizens' Track" to connect grassroots public dialogue to formal governance processes.

Why Is Participation Not Translating Into Real Power?

The summit's most striking finding was a diagnosis that participants across different sectors agreed on: multilateral AI governance has created platforms for exchange but has yet to deliver transparency about how stakeholder input actually shapes outcomes. In other words, people are invited to the table, but there's little visibility into whether their voices influence final decisions.

Participants identified deep asymmetries that undermine genuine participation. AI startups depend on Big Tech companies for access to foundational models and computing power. Institutions in the Global South depend on actors in the Global North for infrastructure and safety expertise. These power imbalances mean that some voices carry far more weight than others in shaping AI governance frameworks.

The convening's central message was clear: the intersessional period of the UNGDAI should be used to move from diagnosis and dialogue to concrete mechanisms and outcomes, building on the actors and evidence already in place.

How Can AI Governance Move From Talk to Action?

• Embed Human Rights Due Diligence: A new OHCHR B-Tech and BSR workstream is embedding human rights due diligence directly into AI product design, moving beyond policy documents to operational practices.
• Invest in Language Equity and Local Infrastructure: Participants called for sustained investment in language equity, judicial and institutional capacity, and local infrastructure for public participation, ensuring that governance processes are accessible to communities beyond English-speaking Global North institutions.
• Ground Governance in International Human Rights Law: Despite differing entry points, participants converged on the need for international human rights law and the UN Guiding Principles on Business and Human Rights to remain the foundation for all governance efforts.
• Create Accountability Mechanisms: The convening emphasized the need for transparency about how stakeholder input shapes outcomes, moving beyond participation metrics to genuine accountability.

What Concrete Deliverables Are Coming?

The Geneva convening is not just a talking shop. Participants committed to concrete outputs. A practical human rights due diligence guide is planned for early 2027, providing organizations with step-by-step frameworks for embedding human rights considerations into AI development and deployment. The Global South Network for Trustworthy AI is preparing the first full edition of its Global South AI Safety Report, expected in time for the 2027 Swiss AI Summit.

These deliverables signal a shift from abstract principles to actionable tools. The human rights due diligence guide, in particular, addresses a gap that has plagued AI governance: most frameworks describe what should happen, but few explain how organizations actually implement human rights protections in practice.

The convening also launched a "Citizens' Track" designed to connect grassroots public dialogue to formal AI governance processes. This is significant because it acknowledges that AI governance cannot be left to policymakers, industry, and experts alone. Communities affected by AI systems need pathways to influence decisions at the formal level.

Why Does This Matter Now?

AI systems are already embedded in critical decisions affecting billions of people: hiring, lending, criminal justice, healthcare, and content moderation. Yet the people most affected by these systems often have no say in how they are built, deployed, or regulated. The Geneva convening exposed this gap and signaled that the next phase of AI governance must prioritize genuine participation over symbolic inclusion.

The emphasis on Global South perspectives and local infrastructure is particularly important. Much of the current AI governance debate is dominated by institutions and voices from wealthy countries. The convening's focus on language equity and institutional capacity in the Global South suggests a recognition that AI governance frameworks designed without input from developing economies will likely fail to address the unique risks and opportunities those regions face.

As AI systems become more powerful and more integrated into everyday life, the question of who gets to shape their governance becomes increasingly urgent. The Geneva convening did not solve this problem, but it clarified what needs to happen next: moving from dialogue to mechanisms, from participation to power, and from Global North-centric frameworks to genuinely multilateral governance.