United Nations Office for Disaster Risk Reduction

First, the Dialogue should help connect broad governance principles with practical applications. It should show how issues such as safety, accountability, human oversight, interoperability and inclusion matter in real-life settings. For instance, AI is already being applied to strengthen disaster risk knowledge and risk-informed decision support (e.g. in early warning systems) intended to save lives and livelihoods. Second, the Dialogue should respond to the realities and needs of countries, especially least developed countries and small island developing states, which often face the highest levels of disaster risks while having the least capacity to shape or govern emerging technologies like AI. A meaningful outcome would recognize that responsible AI governance depends on reliable and inclusive data, institutional readiness, digital infrastructure, sustainable financing, and a country's capacity to govern and sustain the AI-enabled systems. Third, the Dialogue should make clear that national AI strategies and plans need to be developed through an inclusive, whole-of-government approach. Governance and capacity should not be concentrated in one or two ministries alone. Sectors that will be directly affected by AI-enabled decision-making, including disaster management agencies, need to be part of strategy setting, implementation and oversight. Greater emphasis should also be placed on building locally grown AI expertise so that countries can shape, adapt and sustain AI-enabled systems in line with their own priorities.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Protection and promotion of human rights

Safe, secure and trustworthy AI is a priority because AI-supported systems may inform risk assessments, early warning systems, DRR planning and investments. In these contexts, weak safeguards, poor-quality data or misuse can directly affect lives, livelihoods and public assets. Putting greater emphasis on national capacity building in AI is essential because many developing countries, especially those facing high disaster and climate risk, often have limited institutional capacity, data systems, digital infrastructure or technical skills needed to govern and use AI effectively. Targeted support to these countries is needed for more nationally and locally grown AI expertise. Protection and promotion of human rights is central because AI often relies on data that underrepresents, misrepresents or overlooks vulnerable and marginalized groups. Governance must ensure that AI does not deepen existing patterns of exclusion, discrimination or invisibility of those already most at risk. Because decisions in DRR space can be life-saving and time-sensitive, transparency, accountability and human oversight are also critical. AI should not become a black box that reshapes public action without scrutiny. There must be clarity on how systems are used, who is responsible, and how human judgment remains central, especially when uncertainty is high, and the costs of error fall on vulnerable communities.

One important cross-cutting issue is inclusive national governance. At the country level, AI governance discussions are often led by one or two ministries with mandates on digitalization, innovation or telecommunications. This may result in strategies and capacity-building efforts that are too narrow and insufficiently connected to sectoral priorities. National AI governance should involve all relevant sectors, including disaster management agencies, local authorities and other institutions.

In the DRR space, advances in AI are opening real opportunities. This includes using AI-based technologies in improving disaster risk knowledge through better integration and analysis of hazard, exposure, vulnerability and impact data; supporting multi-hazard risk modelling and mapping; strengthening disaster loss and damage data systems; and helping translate complex risk information into more usable inputs for planning, investment and resilience building. There is also growing interest from countries in using AI to help address capacity constraints. In many settings, institutions face limited technical staff, fragmented data and weak analytical capacity. Certain AI-enabled applications are attractive because they can help process large and complex datasets more quickly, automate parts of analysis, and support decision-making without requiring countries to build entirely new systems from scratch. This is particularly relevant in contexts where infrastructure and institutional capacity are limited, but where the demand for better risk information is high. The most significant challenge is ensuring that these gains do not come at the cost of trust, inclusion and national ownership. In the DRR space, AI can easily produce outputs that appear authoritative even when the underlying data are partial, outdated or biased. This can create false confidence in risk assessments, obscure uncertainty, or sideline local knowledge and institutional judgment. There is also a risk that countries become dependent on external tools, models or providers without sufficient control over data, assumptions or long-term maintenance.

As the first Global Dialogue on AI Governance, it can help build a clearer and more shared understanding of what "AI governance" means in practice, and where international cooperation is most needed. It can help focus attention on areas such as capacity-building, technical assistance, data governance and safeguards for countries with more limited resources and capacities. This is particularly important in contexts where public decision-making has direct implications for people's lives and livelihoods. The Dialogue can also serve as a bridge across existing processes and actors, making it easier for governments, the UN system, scientific communities, civil society and the private sector to share lessons, align efforts and reduce fragmentation. In that sense, it can strengthen international cooperation by supporting more coordinated, inclusive and equitable approaches to governing AI across regions and sectors.

It should connect with existing coordination mechanisms such as UN Inter-Agency Working Group on AI and Working Group on Digital Technologies, as well as build on the UNESCO Recommendation on the Ethics of AI. It should also connect with SDGs and sectoral initiatives where governance issues already arise in practice such as the Sendai Framework for Disaster Risk Reduction and its wider implementation ecosystem. The AI governance agenda has implications for countries' ability to accelerate progress towards their targets by 2030. The added value of the Dialogue is that it can bring these strands together in one universal UN forum. It can help reduce fragmentation, improve visibility of existing efforts, and identify where international cooperation is most needed, particularly on capacity-building, safeguards, data governance and support for developing countries, including LDCs and SIDS.

It should connect with existing coordination mechanisms such as UN Inter-Agency Working Group on AI and Working Group on Digital Technologies, as well as build on the UNESCO Recommendation on the Ethics of AI. It should also connect with SDGs and sectoral initiatives where governance issues already arise in practice such as the Sendai Framework for Disaster Risk Reduction and its wider implementation ecosystem. The AI governance agenda has implications for countries' ability to accelerate progress towards their targets by 2030. The added value of the Dialogue is that it can bring these strands together in one universal UN forum. It can help reduce fragmentation, improve visibility of existing efforts, and identify where international cooperation is most needed, particularly on capacity-building, safeguards, data governance and support for developing countries, including LDCs and SIDS.

Voices that remain underrepresented in global AI governance discussions include developing countries especially LDCs and SIDS and countries in high-risk or crisis-affected settings even though they are affected by gaps in technological access and data. Inclusion requires more targeted support for participation beyond open invitations to join such as targeted briefings and support for written submissions, preparatory sessions to gather views in advance.

It would also be useful to create space for more informal and solution-oriented engagement alongside the plenary sessions. For example, small multistakeholder consultations, or focused expert dialogues could allow participants to engage more directly on implementation questions, especially around capacity-building, safeguards and governance gaps.

A practical example is the shift towards an open, interoperable and AI-enabled risk information ecosystem, as reflected in UNDRR's broader strategic direction. This approach is useful because it frames AI and digital technologies as part of strengthening disaster risk knowledge, improving the accessibility and usability of risk information, and supporting risk-informed policymaking, planning, budgeting and investment. Closely linked to this is the emphasis in UNDRR's Data Strategy on data quality, interoperability, inclusion and responsible governance of risk information. This is directly relevant to AI governance, since the value of AI in disaster risk reduction depends fundamentally on data quality, stewardship and the ability to turn complex information into sound public decision-making.