A global weather conference drew 800 participants from 117 countries to examine AI, trust and warning communication, highlighting the need for inclusive approaches and stronger governance to keep systems reliable.
The third edition of the Weather and Society Conference was held online from 23 to 26 February 2026 under the theme Whole of Society Approaches for Weather-ready Communities. Organized by the World Weather Research Programme's Working Group on Societal and Economic Research Applications (SERA), the conference attracted over 800 registrations from 117 countries, including representatives from national hydrometeorological services (NMHSs), businesses, media, and civil society organizations. The free conference was designed to be accessible worldwide, with sessions scheduled across different time zones.
In her opening remarks, Dr Véronique Bouchet, Senior Director of the Department of Science, Services and Capacity Development at the World Meteorological Organization (WMO), stressed the need for more systemic approaches to warnings, ensuring that critical information reaches those who need it most. "Interdisciplinary perspectives should not be treated as an afterthought, but as a starting point for building effective, system-wide solutions", she said.
AI, trust and the future of weather warnings
Artificial intelligence is no longer a horizon technology for the weather and climate community. It is already reshaping how weather forecasts are produced, communicated and used, from machine learning models that support hazard prediction to tools that classify surface conditions and inform decision-making.
AI-generated content is accelerating the speed at which information spreads, but also the speed at which misinformation does. Deepfakes, synthetic media, and algorithmic narratives could make it harder for the public to distinguish trusted warnings from unreliable information during extreme weather events.
Much of the discussion around AI has focused on outputs: ensuring tools are accurate, ethical, and robust. Participants stressed, however, that the bigger challenge lies in the process itself. What governance structures, transparency mechanisms, and accountability frameworks should guide the development of AI? As noted, trustworthiness is not an inherent property of a model, but is shaped by how it is developed, used, and governed. This raises questions about who defines the standards for ethical AI in weather services, and whose perspectives are included in that process.
As AI becomes more widely integrated in early warning systems and public communication, questions of governance and transparency are likely to become increasingly important. Ensuring that these systems remain reliable and trusted will be a key challenge for weather services in the years ahead.
Communicating warnings that people understand
Lessons from Hurricane Melissa pointed to five principles for effective warning communication: timeliness, timing, technology, tone, and transparency. Examples from South Asia and the Caribbean reinforced these principles, showing that culturally sensitive and unified messaging can make a measurable difference.
Discussions also highlighted the challenges of understanding how the public interprets warnings. How a question is asked can shape the answer. Survey design is not a neutral methodological choice: anchor effects, cross-cultural adaptation challenges, and the gap between stated preferences and actual understanding can all influence the results. Weather services may therefore risk building policy on measurement artefacts rather than genuine public insight.
Who shapes weather knowledge
A recurring theme during the conference was the relationship between scientific systems and Indigenous and local knowledge. These systems often validate each other, yet they remain institutionally separated. Cases where local observations aligned with ground-level data but diverged from climate models suggest that more transdisciplinary approaches may be both ethically desirable and technically valuable.
Discussions about knowledge also raised broader questions about inclusion and participation. Inclusion is often framed as a communication or access problem, but the evidence points to something deeper. Whether the context was women in storm chasing, disabled communities in Nepal, or children in Indonesia, marginalized groups remain recipients of climate information rather than co-producers of it. Genuine inclusion requires a shift in who holds decision-making power, not just who receives the message.
