World Meteorological Organization

For WMO and the meteorological, hydrological and climate community, success would mean, 1. A common framework affirming that AI development must remain open, transparent and interoperable across national boundaries. Earth system data –the foundation of weather and climate services –must continue to flow freely in accordance with the WMO Unified Policy for the international exchange of earth system data and the FAIR Data Principles. WMO's community has progressively built consensus around this vision, from the Abu Dhabi Conference (Sep 2025) through the Extraordinary Session of the World Meteorological Congress (Oct 2025), and the Dialogue should consolidate and globally endorse it. 2. Explicit recognition of the digital divide as a structural governance challenge. Governance frameworks that do not address this gap risk codifying existing inequalities. The Dialogue should produce concrete commitments to technology transfer, shared infrastructure, and capacity development. 3. Affirmation that new prediction technologies complement, rather than replace, national institutions which bear legal and public accountability for forecasts and early warnings. Trust in forecasts depends on systems that are auditable, verifiable, and subject to rigorous human oversight. 4. Recognition of the need for inclusive mechanisms for multi-stakeholder coordination spanning public, private, and academic actors. WMO has created the Joint Advisory Group on Artificial Intelligence (JAG-AI) and calls on the Dialogue to establish equivalent mechanisms.

• Open-source software, open data and open AI models
• Safe, secure and trustworthy AI
• AI capacity-building

Safe, secure and trustworthy AI - This priority is foundational to WMO's mandate. Weather, climate, and hydrological forecasts directly inform decisions about the protection of life and property, managing water resources, food systems, and evacuation plans. Systems operating in this domain must be scientifically sound, rigorously verified and operationally reliable. Opaque or unverified models risk eroding the public trust that official warnings depend on. AI capacity building - The countries that are most vulnerable to weather, climate and hydrological events - LDCs and SIDS - are also those with the least access to computing infrastructure, data, and technical expertise. Yet, they could benefit the most from access to these technologies to leapfrog their capacity. Capacity development must meet Members at their current levels - through training programmes, access to local data, use cases and open collaboration. Open-source software, open data and open AI-models: Open-source approaches lower the barriers to access and innovation for developing countries and ensure transparency and trust in AI applications and their use. In the context of capacity development for meteorological services and early warning systems, open-source approaches are more likely to foster a sustainable and long-lasting legacy of the capacity established through initial investment in countries.

Guardrails for foundational observational and reanalysis data: The data underpinning weather, climate, and hydrological AI prediction activities are collected and maintained at substantial public cost. For this reason, significant gaps remain in foundational data, particularly in the Global South. As all actors use this data to train AI prediction models, there needs to be recognition of the importance of foundational observational and reanalysis data, particularly in an era when historical climate data will likely not reflect our future climate reality. Governance frameworks should recognize the importance of underpinning data and consider ways to protect and strengthen its existence. Accountability when it fails: Application of Ai in the context of public safety requires careful attention on authoritative role and accountability. Governance frameworks need to address accountability to ensure legal responsibilities are defined when weather, climate, and hydrological predictions and warnings fail.

Regarding the thematic areas selected above, including (1) Safe, secure and trustworthy AI; (2) AI capacity-building; and (3) Open-source software, open data and open AI models, we are seeing several challenges emerge across the weather, climate and water science and services. In particular, inequity is a major challenge. AI's impact depends on national capability to govern, integrate, and operate new tools, but not all countries have the same level of capacity to do so. As a result, inequity can further widen the digital divide. Another challenge is fragmented standards. Common standards and safeguards are essential for scientifically sound, transparent and reliable AI systems and to reduce risks, including opaque models, uneven performance and fragmented service delivery. Finally, trust and accountability are another major challenge. Building and maintaining public trust is essential for the adoption, collaboration, and sustainability of AI for weather, climate, and hydrological forecasts. It depends on credible, transparent, and scientifically rigorous practices and responsible innovation. However, several opportunities also exist. For example, AI represents a powerful and transformational enabling technology that can significantly democratize access to advanced weather forecasting, climate prediction, water management and disaster risk reduction in lower-income countries, including faster and more reliable early warning systems. By lowering barriers to participation, more Members and partners can contribute their data, expertise and innovations to a shared global ecosystem.

The AI Dialogue can consolidate and endorse the principles developed by the WMO community, including the WMO Unified Data Policy (https://wmo.int/wmo-unified-data-policy-resolution-res1), which promotes the international exchange of earth system data and the FAIR Data Principles. Sustainable AI for environmental services depends on high quality data and open tools and interfaces. Promoting open data sharing, including with public and private stakeholders, can make datasets widely accessible and fit for purpose enabling equitable access to AI innovation and ensuring that all countries, especially those with limited infrastructure, can participate in and benefit from AI advancements. Additionally, the call can endorse the WMO Call to All Stakeholders to Collaborate on the Development of AI and Machine Learning Environmental Monitoring, Prediction Technologies, Tools and Applications (https://meetings.wmo.int/cg-ext-2025/_layouts/15/WopiFrame.aspx?sourcedoc={5C451393-60CF-4A5A-83E2-64564ADDB4BA}&file=Cg-Ext(2025)-d02-3(2)-WMO-CALL-DEVELOPMENT-AI-ML-TECHNOLOGIES-approved_en.docx&action=default). This is a call for collaborative action of public, private and academic sectors to meet the need for sustainable development, the reduction of loss of life and property caused by natural hazards and other catastrophic events related to weather, climate, hydrological, marine and related environmental events, as forged by the United Nations Early Warnings for All initiative, by enhancing and leveraging the application of AI/ML technologies to environmental prediction systems and service delivery, including early warnings. And finally, the WMO AI Conference statement (https://wmo.int/sites/default/files/202509/Conference%20Statement%20180925.pdf), which presents the shared position and vision of the WMO weather, climate, and water scientific community. The Dialogue can also advance a framework for public-private data relationships. Recognizing that AI is built on the availability and accessibility of data, public-private data sharing is essential to make datasets widely accessible and fit for purpose to fully realize and scale the opportunities afforded by AI.

Voices, communities and perspectives from developing countries are most likely to be absent due to limited technological capacity and interpretation challenges. As a result, the Dialogue risks further amplifying the digital divide. However, real time interpretation can help facilitate higher levels of engagement.

In September 2025, WMO and the United Arab Emirates National Center of Meteorology convened the WMI AI Conference in Abu Dabi, UAE. This conference brought together experts and representatives from across research institutions, national meteorological agencies and industry, each bringing their own unique capabilities, to discuss opportunities to build an AI ecosystem for global good. The conference is a good example of a platform that brought experts together to discuss challenges and concrete solutions. Outcomes from the conference are summarized in the WMO AI Conference statement (https://wmo.int/sites/default/files/202509/Conference%20Statement%20180925.pdf)