Diverse AI Solutions | AI2030

Success for this first Dialogue for me is about proving that genuine multilateral cooperation on AI is actually possible. From where I sit in the private sector, four things would make this a real success: 1. A framework we can actually build to. Not a declaration, but a starting point. Needs to include shared principles with a clear process for keeping them current as the technology evolves. Certainty, even provisional certainty, enables responsible innovation. 2. A table where everyone is actually shaping the outcome. Industry, civil society, researchers, governments (not consulting in silos, but genuinely co-authoring the norms we'll all be held to). UNESCO has the credibility to make that happen. 3. A commitment that the benefits reach everyone. AI governance has to actively ask who is being left out. The communities with the least representation in these conversations are often the most exposed to AI's consequences. Success means building that reality into the framework from the start, not as an afterthought. 4. Less fragmentation. The current patchwork of national and regional frameworks isn't protecting anyone well. It's creating confusion. This Dialogue can start to make sense of that landscape: where can we harmonize, and where should we respect regional difference? Five years from now, we'd love to look back at Geneva as the moment AI governance stopped being a conversation and started being infrastructure that is built for everyone, not just those already at the table.

• Safe, secure and trustworthy AI
• Protection and promotion of human rights

1. AI Capacity-Building AI's benefits won't be evenly distributed unless we deliberately invest in making them accessible. For the private sector, this is about ensuring the next wave of innovation comes from everywhere, not just a handful of well-resourced markets. Talent, infrastructure, and know-how need to travel further and faster. 2. Safe, Secure and Trustworthy AI This is the foundation everything else rests on. Without it, public trust erodes and adoption will not move forward. This is bad for everyone. Industry has a direct stake in getting safety and security right, not because regulators demand it, but because products that people can't trust don't have a future. 3. Protection and Promotion of Human Rights Technology built without human rights at the center tends to replicate and amplify existing inequalities. As builders and deployers of AI systems, the private sector has a real responsibility here and an interest in clear, consistent standards so that "doing the right thing" isn't a competitive disadvantage.

From our vantage point as an AI consulting firm in North America, the governance gaps are showing up in practical ways every day. On capacity-building, the knowledge gap is becoming a power gap. Organizations without resources to navigate AI (smaller companies, public agencies, under-resourced communities) are being left behind or exposed. On safe and trustworthy AI, our clients are navigating a patchwork of frameworks that don't talk to each other. That inconsistency doesn't just slow things down, it creates openings for bad actors and erodes public trust. On human rights, we're seeing systems deployed in hiring, healthcare, and public services without adequate safeguards. The governance infrastructure simply isn't keeping pace with deployment. The common thread: the technology is moving faster than the guardrails, and the communities with the least voice are absorbing the most risk.

The AI Dialogue has a role no single country or regional body can fill. It can be the place where fragmented efforts start to come together. Right now, international cooperation on AI governance is more aspiration than reality. Frameworks exist, but they don't connect. Conversations happen, but in parallel, not together. The Dialogue can change that by doing three things: 1. Being a clearing house for what's working. Countries and sectors are running governance experiments right now. The Dialogue can surface what's actually effective and help good ideas travel faster. 2. Building trust between actors who don't naturally talk. Governments, industry, civil society, and underrepresented communities all need to be in the room — not sequentially, but together. The Dialogue is one of the few spaces with the legitimacy to make that happen. 3. Holding the long game. AI governance isn't a problem you solve once. The Dialogue can establish the rhythm (regular, structured, inclusive) that keeps cooperation alive as the technology evolves. The opportunity here isn't to produce the definitive global AI framework. It's to make sustained, meaningful cooperation a habit.

For the Dialogue to work, it needs to be designed for real participation, not just representation. Governments set the conditions for cooperation. Industry brings technical knowledge and deployment experience, but needs to show up as a governance partner, not a lobbyist. Civil society and affected communities have irreplaceable proximity to AI's real-world harms and need genuine influence, not just a seat at the table. Researchers serve as honest brokers between technical complexity and policy. A few format recommendations: 1. Mix plenary with working sessions. Big room moments build shared understanding; smaller groups are where progress actually happens. 2. Build in follow-up. A Dialogue without accountability becomes a talking shop. Center Global South voices as co-authors of governance, not recipients of it. 3. Design for iteration. This should be the first in a series, not a one-time event. The measure of a good format is simple: does every stakeholder leave feeling they shaped something real?

The communities most affected by AI are often the least represented in conversations about how it should be governed. That includes people in the Global South, indigenous communities whose data and languages are used without consent, low-income populations navigating AI in healthcare and social services, and frontline workers whose jobs are being transformed with little say in how. Governance designed without them will have blind spots that show up as harms. A few ways to close the gap: 1. Fund participation, don't just invite it. Translation, travel, and capacity support should be built into the Dialogue's budget. 2. Go beyond capital cities. Regional consultations can reach communities that will never make it to Geneva. 3. Create structured input mechanisms that ensure marginalized perspectives actually shape outcomes, not just inform them.

Traditional conference formats, panels and plenaries, tend to generate discussion but not decisions. For a Dialogue that actually moves things forward, a few formats worth looking into: 1. Scenario-based working sessions. Put stakeholders in front of real governance dilemmas and work through them together. 2. Structured stakeholder exchanges. Pair industry representatives directly with civil society or Global South counterparts to develop joint positions before plenary sessions. 3. Live gap-mapping. Use sessions to collectively identify where existing frameworks conflict, overlap, or fall short. 4. Open feedback loops. Simple, accessible mechanisms during and after the Dialogue for communities who weren't in the room to respond to what came out of it.

A few that stand out: 1. Community-led impact assessments. When communities define the questions, particularly in housing, healthcare, and public services, the findings are harder to ignore and the outcomes are better. 2. Participatory design in practice. Bringing affected communities into the design process, not just consultation after the fact, produces fewer harms and builds the kind of trust that makes adoption more durable. 3. Peer-to-peer capacity building. Grassroots networks training community advocates to understand and challenge AI systems in their own contexts have proven more effective than top-down literacy campaigns. The common thread is ownership. Governance works when the people most affected by it helped build it.