From Participation to Agenda-Setting: lessons for the next global AI Summit

The recent global discussions around artificial intelligence governance have revealed both the urgency of the topic and the limitations of the processes currently shaping it, as we have witnessed with the AI Impact Summit in New Delhi. As preparations move toward future milestones, particularly the planned Swiss AI Summit in 2027, it is essential to reflect on what worked, what did not, and how global governance processes can evolve.

As defended by the Global South Alliance, one valuable reference point is the experience of Internet governance, especially the process surrounding NetMundial (2014) and NetMundial+10 (2024). These initiatives demonstrated both the promise and the complexity of building truly participatory global governance frameworks. The AI governance community would benefit greatly from revisiting these lessons.

The AI Impact Summit’s Missed Opportunities

The AI Impact Summit in New Delhi generated significant anticipation, primarily because it marked the first edition to be held in the Global South. This sense of anticipation was reinforced by India’s presidency of BRICS, which followed the group’s adoption of a Statement on the Global Governance of Artificial Intelligence. 

In contrast to earlier AI summits, characterized by low institutionalization and few tangible outcomes, this edition was seen as an opportunity to elevate the priorities of developing countries, strengthen South-South coordination, and forge clearer connections with ongoing multilateral initiatives, including the UN International Scientific Panel and the UN Global Dialogue on AI. 

However, the event did not fulfill those hopes. Rather than advancing the ethical and political dimensions of AI, the Summit’s general tone leaned heavily toward industrial policy and innovation. The event’s structure provided major technology companies with a high-level platform, while civil society and rights-based debates remained largely on the margins of the official discussions and media coverage. Calls for more substantive discussion on the global governance of AI were voiced by civil society and a handful of other actors, notably Brazilian President Lula da Silva. 

In the end, the Summit’s outcome declaration fell short of the expectations voiced by actors from the Global South, reinforcing the pattern established in previous editions and highlighting the fragmentation of AI governance. The final text offers a principles-based statement, but lacks concrete commitments or avenues for effective multilateral action. While it defends a multistakeholder approach, in practice this framework prioritizes collaboration between governments and tech corporations, rather than establishing mechanisms for broad-based societal participation and accountability.

Meaningful Participation: lessons from Internet Governance

One of the most important lessons from Internet governance processes is that participation must be meaningful, not symbolic. Over the past two decades, Internet governance has developed a set of practices designed to ensure broad stakeholder engagement –  including governments, civil society, academia, the private sector, and technical communities – consistent with the multistakeholder vision articulated during the World Summit on the Information Society (2003; 2005). 

Processes such as NetMundial created open consultation mechanisms, transparent documentation, and opportunities for diverse actors to shape the agenda. One insight from NetMundial+10 deserves particular attention: “Documentation of how contributions were made, evaluated, and incorporated into the process is as important as the documentation related to dissenting and divergent views.”

Transparency is not only about publishing final outcomes. It is about documenting the path taken to reach them: who contributed, which ideas were accepted or rejected, and why. Without this visibility, participation risks becoming performative rather than substantive. 

As discussions move toward the Swiss AI Summit in 2027, organizers have an opportunity to apply these principles from the outset. This includes:

• Creating open consultation processes early in the agenda-setting phase;
• Clearly documenting how stakeholder input influences outcomes;
• Ensuring that dissenting perspectives are recorded and respected;
• Designing governance mechanisms, with efforts to uphold the principle of meaningful participation, that go beyond consultation and enable co-creation.

AI governance will not succeed if it replicates closed or opaque diplomatic processes. Trade negotiations should not be the template for AI governance. The legitimacy of global AI governance will depend heavily on how inclusive and transparent the process is, not only on the final policy recommendations.

Inclusion vs Agenda-Setting

Another challenge that emerged in recent global AI discussions is the difference between being included for visibility and being invited to shape the agenda. Many actors, including those from the Global South, are increasingly present in global forums. This was the case in New Delhi, where we could observe the presence of local non-governmental organizations in the official program of the Summit, although in a limited capacity. Yet presence alone does not necessarily translate into influence.

This distinction lies at the heart of the spirit behind initiatives such as the Global South Alliance. The concern is not simply about representation, but about participatory power: the ability to define priorities, frame debates, and influence outcomes. Civil society organizations do not want to be part of a “Global South quota” for the sake of equity. It doesn’t matter if you’re invited to be in the living room photo if the important decisions are made in the kitchen, as one leader of a civil society organization mentioned in a side-event in New Delhi.

In some recent processes, including the AI Summits and initiatives such as the UN’s Global Digital Compact, civil society and participants from the Global South faced structural barriers, such as (i) limited access to preparatory documents, (ii) insufficient transparency about decision-making processes, (iii) short timelines for consultation and engagement. These limitations had real consequences. Many voices that could have contributed meaningfully to the discussions were unable to do so.

In some cases, actors even hesitated to express criticism publicly. For example, some organizations were concerned that publishing open letters or raising procedural concerns might negatively affect visa processes or their ability to attend international meetings. This situation illustrates a broader governance challenge. A truly democratic global governance process must ensure that participants can engage without fear of procedural or political repercussions.

A moment of public visibility?

At the same time, something remarkable happened during the recent summit in India: public visibility increased significantly. For many people, this was the first time they had heard about a global AI summit. In countries such as India and Brazil, discussions about the summit appeared frequently on television and in the media. The topic reached audiences far beyond the traditional policy and technology communities.

It is true that the media narrative focused on major figures in the corporate world, such as the non-shake of hands between Sam Altman (OpenAI) and Dario Amodei (Anthropic). In India, much of the media coverage focused on Modi’s major industrial policy announcements and very little on what organized civil society had to say.

This moment of public attention creates a unique opportunity. Rather than allowing this momentum to dissipate, the global community should build on it by creating accessible materials that help broader audiences understand what is at stake. This includes:

• Explainers about AI governance processes and the role of the AI Summits;
• Educational materials for universities and research institutions on how to contact the Co-Chairs of the Summit and voice their concerns;
• Resources for civil society organizations and policymakers, supporting organizations outside the “digital rights bubble”;

The scientific community, in particular, has an important role to play, adding to the efforts of civil society. Researchers can help translate complex technical debates into concepts that policymakers and the public can engage with. Initiatives such as AI Safety Commons illustrate both the promise and the conceptual challenges ahead. For example, when we talk about “AI safety,” what exactly do we mean?

Are we primarily addressing “accidental harms”, such as system failures or unintended consequences? Or are we also addressing intentional harms, including misuse of data, manipulative design, or weaponization?

These distinctions matter because they shape how governance frameworks are designed. They influence which actors are responsible, what mitigation strategies are prioritized, and how accountability mechanisms are structured. In other words, part of the challenge ahead is not only technical or regulatory, but it is also conceptual. Without clear concepts, we cannot aim for political commitments and normative goals.

The commitments of the next Summit should not be focused on the number of people to be trained in AI or the amount of billions of dollars to be invested. Civil society demands clearer concepts about what fundamental rights are necessary, what we understand by existential risks, what we understand by AI safety, and what we understand by limits to the anthropomorphization of these systems.

Toward a more democratic AI Summit?

The next phase of global AI governance will require more than technical expertise or diplomatic negotiation. It will require process innovation (or it will require what Albert Hirschmann called possibilism, meaning finding hidden opportunities for progress in seemingly intractable, complex situations, often by rejecting rigid theoretical models in favor of interdisciplinary approaches). Internet governance has already experimented with models of multistakeholder participation, open consultation, and transparent documentation. While imperfect, these experiments provide valuable guidance for the AI governance community.

Since the original conception of Internet Governance within the WSIS framework, a lot has changed. This has been recognized in NetMundial+10, which set out to address not only Internet Governance, but digital policy processes as a whole, as well as in the WSIS+20 review (2025), which highlighted a human rights approach, coupled with the defense of multistakeholderism and the coordination among processes.

The deeper question that we face in a “post-New Delhi momentum” is how to build governance processes that are legitimate, inclusive, and trusted across regions and communities, if we consider – even hypothetically – that “AI Summits” are part of the AI governance arena and have an important role to play.

The upcoming Swiss AI Summit in 2027 could mark an important step toward a more democratic model of global AI governance if it integrates these lessons: ensuring meaningful participation, enabling agenda-setting from diverse actors, leveraging public visibility, creating clear follow-up mechanisms and avoiding duplication. This opportunity could be used to apply the NetMundial+10 guidelines in practice, setting a strong precedent for multilateral processes, while simultaneously improving coordination among the implementation spaces of the Global Digital Compact (GDC), where AI stands as an objective.

Governments with autocratic intentions will attempt to undermine the multilateralism of the UN system and try to minimize the democratic importance of multi-sectoralism. With the militarization of AIs and increasing geopolitical conflicts, there will be efforts to ensure that the Summit only deals with trade issues and “Pax Silica”. The Swiss government will have a duty to confront these regressive forces if it wants to advance truly democratic approaches.

There is a body of knowledge built over decades, forged in the democratic experimentation of technical communities and organized civil society in Internet governance, that cannot be wasted. AI Summits can become more legitimate new institutional arenas if they are able to seriously take a local/global and multistakeholder approach. This requires an intention of asymmetrical trade-offs, so to speak. Faced with multi-billion dollar corporations with close ties to Departments of Defense (where the power of capital merges with military power), democratic countries must intentionally seek to actively listen to public interest entities. Without this democratic intentionality, the Summits will continue to be merely a demonstration of industrial policies and corporate power.