As discussions on global AI governance accelerate, the United Nations has emerged as one of the few institutions with the political legitimacy to convene governments, civil society, academia, and the private sector around a shared agenda. In a fragmented geopolitical landscape increasingly shaped by competing national AI strategies and corporate concentration (as seen in past “AI Summits”), the UN’s effort to create the Global Dialogue on AI is both necessary and timely.

The initiative represents an important opportunity to move beyond isolated national approaches and toward a more democratic conversation on the global implications of artificial intelligence. For many actors from the Global South, civil society, and marginalized communities, the UN remains one of the only spaces where participation in shaping digital governance norms is still possible. Yet legitimacy cannot rely solely on institutional mandate. It also depends on process.

Recent online consultations organized under the Global Dialogue on AI have revealed important procedural shortcomings that risk undermining the very principles of inclusion and participation the initiative seeks to promote. Participation numbers alone do not necessarily translate into meaningful inclusion, particularly when transparency and accountability mechanisms are absent. By failing to create opportunities for diverse stakeholders, particularly from the Global South, not only to participate, but to shape agendas and influence outcomes, the Dialogue risks replicating mistakes from previous forums.

The scale of engagement is undoubtedly impressive. One consultation gathered more than 600 registered participants, while another reached around 300. But these numbers also expose a structural problem: it is practically impossible for the Secretariat, organizers, and co-chairs to meaningfully hear every voice in such compressed formats, with 3 minutes for each speaker. Long hours of consecutive interventions are not only exhausting for participants, but also likely for the organizers themselves.

The consultation held on May 12 made these tensions especially visible. Many individuals and organizations who had registered to speak never appeared on the speakers’ list, despite being present and prepared to contribute. This was particularly frustrating for civil society stakeholders, many of whom had already experienced similar situations during the Global Digital Compact consultation process, where restricted participation formats also became a recurring pattern. The absence of publicly available criteria explaining how speakers were selected immediately raised concerns among participants. Who was prioritized? Why were some voices heard while others were excluded? Were regional, sectoral, or gender considerations taken into account? Without transparency, these questions remain unanswered.

This is not a minor procedural detail. A process that presents itself as global, participatory, and inclusive cannot afford opacity in determining whose voices enter the conversation. When participants cannot understand how decisions are made, trust erodes, and with it, the legitimacy of the process itself.

Of course, organizing genuinely democratic global consultations is extraordinarily complex. No multilateral process can realistically accommodate hundreds of interventions in a fully open format without facing logistical constraints. But acknowledging complexity should not become an excuse for avoiding transparency. On the contrary, complexity makes transparent governance even more necessary. There are practical measures that could significantly improve the process moving forward.

The Global Dialogue on AI should establish and publicly communicate transparent criteria for speaker selection, drawing on examples such as NETMundial+10, which also faced the challenge of organizing broad and diverse participation. These criteria could prioritize regional diversity, gender balance, stakeholder representation, and inclusion of underrepresented communities. Such mechanisms would not eliminate all tensions, but they would at least provide predictability, accountability, and procedural legitimacy. Participants may accept limitations more readily when they understand the rationale behind decisions.

The process should also encourage and prioritize submissions from coalitions and collective statements. In global governance spaces, coalitions often represent dozens of organizations and communities. Allowing coordinated interventions would make consultations more manageable while also ensuring broader representation in less time. Collective contributions are not a reduction of participation; they are often a more efficient expression of democratic consensus-building. The WSIS+20 process held in 2025 was a good example of this.

The challenge facing the Global Dialogue on AI is therefore not whether participation exists, but whether participation is being meaningfully structured in ways that preserve legitimacy, fairness, and trust. At a moment when AI governance is becoming increasingly central to global politics, the UN cannot simply reproduce procedural opacity while claiming inclusiveness. If the Global Dialogue on AI is to become a credible pillar of global AI governance, its commitment to participation must extend beyond open registration forms and into the governance architecture of the process itself. 

Data Privacy Brasil will continue its efforts in this process together with international coalitions: the Global South Alliance, the Global Digital Rights Coalition and the Global Digital Justice Forum.

Veja também

Veja Também