DTH-Lab

A successful first Global Dialogue on AI Governance should create real momentum for coordinated global action - not only high-level statements. First, it should achieve initial consensus on core principles such as safety, security, human rights, transparency, accountability, and human oversight (i.e. building on approaches like the EU AI Act). Even partial alignment would help reduce fragmentation across regulatory frameworks. Finally, success would include a shared commitment that AI must serve the public interest - not only by mitigating risks, but by actively contributing to societal well-being.

• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Placing human rights at the center ensures a normative anchor that is globally recognized and helps avoid fragmented or purely market-driven approaches. This is particularly important in high-impact areas like healthcare or public administration, where consequences are systemic. At the same time, considering the broader societal implications of AI highlights that governance cannot be limited to technical risk management. AI reshapes access, participation, and power structures, and these effects need to be explicitly addressed. Overall, the priorities reflect a shift from "what should be done" to how governance can be implemented in a way that builds trust and delivers public value - including through approaches like the EU AI Act that aim to translate principles into enforceable frameworks.

Yes. First, power concentration and market structure deserve more explicit attention. The development and deployment of advanced AI is increasingly controlled by a small number of actors, creating structural dependencies and raising concerns about democratic accountability. In extreme cases, this concentration of technological and informational power can enable forms of political control that are difficult to contest, making the intersection of AI and authoritarian or exclusionary governance models a critical issue. Second, public interest governance is not sufficiently foregrounded. Current discussions tend to focus on risk mitigation, but less on how AI can be actively steered toward societal goals. There is a need for governance frameworks that ensure AI contributes to public value - particularly in sectors such as healthcare, education, and public administration. Third, the relationship between AI and broader socio-economic transformation, including debates around degrowth and sustainability, is largely absent. While AI is often framed as a driver of efficiency and growth, its environmental footprint, resource use, and impact on labor and consumption patterns require more critical reflection. At the same time, degrowth-oriented approaches raise important equity concerns: expectations to limit growth or resource use may disproportionately affect low- and middle-income countries that still rely on economic expansion to meet basic needs. AI governance therefore needs to address sustainability in a way that is globally fair and does not reinforce existing inequalities.

With the EU AI Act, Europe has set a global benchmark. Yet a clear governance gap remains in implementation: many institutions lack the technical expertise, data infrastructure, and operational guidance to translate regulation into day-to-day practice. At the same time, fragmentation and dependency persist. Differences in digital maturity across countries and sectors - especially in healthcare, the sector I am working in - lead to uneven application. These gaps, however, create a strategic opportunity. If we succeed in operationalizing trustworthy AI, it can turn governance into a practical advantage and support more human-centered, public-interest-driven AI.

The AI Dialogue can advance international cooperation not by creating immediate global rules, but by building the conditions that make cooperation more realistic over time. In a fragmented governance landscape, its main value is as a bridge between states, regions, science, industry, and civil society, helping develop trust, a more shared language, and clearer priorities. The UN frames it as an inclusive platform to discuss international cooperation, share best practices and lessons learned, and support open and transparent discussions on AI governance. Its more ambitious potential is that it could gradually turn scattered discussions into a more continuous form of cooperation. In that sense, the AI Dialogue is less a rule-making body than a convening architecture: a space to build trust, identify practical convergence, and prepare the ground for more coordinated global AI governance later on.

The AI Dialogue should build on initiatives that already provide key elements of AI governance, rather than duplicating them. This includes the OECD AI Principles or multistakeholder forums such as the Internet Governance Forum. It should also connect closely with the Independent International Scientific Panel on AI, as well as regional and national regulatory efforts and technical standard-setting processes. Its added value would not mainly be new principles, since many already exist. Rather, the AI Dialogue could serve as a coordination mechanism across a fragmented governance landscape. It could help improve interoperability between different frameworks, strengthen links between scientific evidence and political discussion, and include countries and stakeholders that are often less visible in existing AI governance forums. In that sense, its main contribution would be to create more complementarity, continuity, and gradual convergence across existing initiatives, rather than institutional fusion.

Different stakeholders should contribute in complementary ways: governments and regulators can bring policy experience, industry can share deployment realities, academia can provide independent (inter-)disciplinary evidence, ethic experts, human rights experts and civil society can raise rights and equity concerns, and affected communities should be included not only as consultees but as participants. In terms of format, the AI Dialogue should combine plenary sessions with smaller (inter-)disciplinary working groups, written submissions, and a light citizen-assembly element or citizens' panel, so that public values and lived experience are not lost in mainly diplomatic or technical discussions.

Underrepresented voices in global AI governance are often those who make up a large share of the world's population but have the least access to agenda-setting spaces: young people, rural communities, low-income populations, linguistic minorities, Indigenous communities, persons with disabilities, workers affected by automation, and stakeholders from low- and middle-income countries, especially outside major policy and technology hubs. The problem is not only demographic diversity, but also structural diversity: too many discussions still overrepresent governments, large firms, and experts from a small number of countries and institutions. For a truly global process, representation should reflect the social and geographic distribution of the world more seriously. A small citizens' panel or deliberative mini-public could also help bring in public perspectives that are usually missing from diplomatic and technical discussions. My point is that inclusion should not happen only through consultation, but through meaningful participation in agenda-setting, discussion, and follow-up.

As mentioned above, a small citizens' assembly or citizens' panel linked to the AI Dialogue would be my suggestion. This could help bring in public values, lived experience, and perspectives that are often missing from diplomatic or highly technical discussions, making the process more legitimate, grounded, and socially responsive.

A strong example is the EU AI Act, which translates broad AI governance principles into a risk-based legal framework. It distinguishes between prohibited uses, high-risk systems, transparency obligations for certain systems, and lighter rules for lower-risk applications. This is important because it shows one concrete way of moving from abstract principles such as trustworthiness or human oversight into enforceable legal obligations. The EU has also paired the Act with implementation-oriented mechanisms such as the AI Office and the AI Pact, which are meant to support compliance, knowledge sharing, and preparation before the rules fully apply.