The University of Alabama

A successful first Global Dialogue on AI Governance should produce concrete governance priorities, not only broad principles. One important outcome would be recognition that AI governance must include the physical infrastructure behind AI, especially data centres, and their growing energy and water demands. The United Nations has already acknowledged that rising AI energy and water use can impose costs on vulnerable communities, so the Dialogue should move this from concern to action. From my perspective as a hydrologist and fluvial geomorphologist, success would mean three practical advances. First, agreement that AI infrastructure governance should include transparent reporting of water withdrawals, source water type, seasonal demand, and local basin stress, rather than relying only on annual corporate averages. Second, the Dialogue should encourage governance approaches that protect communities and ecosystems in water stressed regions by considering environmental flows, drought conditions, and cumulative local impacts. Third, it should promote interoperable reporting and assessment frameworks so countries can compare risks consistently while adapting them to local contexts. The Dialogue will be most successful if it makes clear that responsible AI is not only about models and safety. It is also about where AI is built, how resources are used, and whether the benefits of AI are achieved without transferring environmental costs to already vulnerable people and places.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

I selected these priorities because AI governance should not focus only on algorithms, model safety, and digital harms. It should also address the physical systems that make AI possible, especially data centres and their growing demands for water and energy. My perspective comes from hydrology and fluvial geomorphology, where water use is understood in relation to river basins, seasonal variability, drought conditions, and ecosystem limits. The social, economic, ethical, and technical implications of AI are important because data centre water use can create real tradeoffs between technological growth, community water security, and environmental sustainability. Transparency, accountability, and human oversight are essential because annual corporate averages often hide when and where water stress is most severe. Meaningful oversight requires clear reporting of water withdrawals, source water, timing of demand, and local basin conditions. Protection and promotion of human rights matters because vulnerable communities should not bear the environmental costs of AI infrastructure without visibility, participation, and safeguards. Access to water, environmental quality, and fair resource allocation are directly relevant in regions already facing drought, water scarcity, or weak infrastructure. Interoperability of governance approaches is also necessary because countries need reporting and assessment frameworks that are comparable across jurisdictions but still flexible enough to reflect local hydrologic realities. Good AI governance should be able to connect technical innovation with basin scale environmental limits and social accountability.

Yes. One major cross cutting issue is the physical footprint of AI infrastructure, especially its water, energy, and land demands. This issue is often treated as secondary, yet it is central to whether AI development is sustainable and equitable in practice. AI systems do not operate in abstraction. They depend on data centres, cooling systems, electricity networks, and water supplies that are embedded in specific places. A key emerging gap is that water use is usually discussed in aggregate terms, even though the real risk is local and seasonal. The same amount of water use can have very different consequences depending on basin stress, drought conditions, environmental flow needs, and competing community demands. Governance discussions should therefore move beyond simple annual efficiency metrics and consider where and when resource demands occur. Another overlooked issue is cumulative impact. A single facility may appear manageable, but multiple facilities within the same watershed can create significant pressure on rivers, aquifers, and local utilities. This is especially important in rapidly growing regions where AI infrastructure may expand faster than water planning and environmental review. Finally, AI governance should pay more attention to infrastructure transparency. Public reporting should distinguish between withdrawals and consumption, identify source water, and clarify whether alternative supplies such as reclaimed water truly reduce local pressure. Without this level of clarity, it is difficult to assess real impacts or design fair governance responses.

Across regions, one of the most significant governance gaps is that AI is still discussed mainly as a digital issue, while the physical infrastructure that supports it remains underexamined. Rapid expansion of data centres is increasing demand for electricity and, in many places, water for cooling. Yet governance remains uneven in how it addresses where water is sourced, when it is used, and how demand interacts with drought, basin stress, ecosystem needs, and competing local uses. This matters globally because hydrologic risk is not uniform. The same level of water use can have very different consequences depending on climate, river basin conditions, infrastructure capacity, and local governance. A facility that appears efficient in annual terms may still create serious pressure during low flow seasons or in already stressed catchments. Many current governance approaches do not adequately distinguish between water withdrawals and water consumption, or require disclosure of source water, timing of use, and cumulative watershed impacts. At the same time, this is a major opportunity. AI governance can become stronger by linking infrastructure planning to basin scale water risk assessment, environmental flow protection, and community level transparency. A more globally coordinated approach could support comparable reporting across jurisdictions while allowing adaptation to local hydrologic realities. This would help ensure that AI development advances innovation without shifting environmental costs onto vulnerable communities or already stressed water systems.

The AI Dialogue can advance international cooperation best by doing what many existing initiatives cannot do on their own: bring governments, technical experts, industry, civil society, and UN system actors into one inclusive forum that is global in reach and not limited to a single region or policy tradition. Its role should not be to duplicate existing principles. It should connect them, identify where they overlap, and translate them into practical areas of cooperation, especially for countries that are not shaping the current governance agenda from the center. From my perspective, the Dialogue is especially valuable if it expands cooperation beyond model level concerns and addresses the physical infrastructure of AI, including data centres, energy demand, and water consumption. This is where international cooperation is badly needed, because water risk is local but the AI industry is global. Shared approaches to transparency, reporting, and impact assessment would help countries compare practices, learn from one another, and avoid shifting environmental costs onto vulnerable communities. The Dialogue can help create common expectations on disclosure of water withdrawals, water consumption, source water, timing of use, and basin level stress. It can also create a bridge between AI governance and sustainable development by linking digital governance to human rights, environmental sustainability, and resource stewardship. That would be a strong contribution of the UN system and a meaningful step toward globally relevant AI governance.

The AI Dialogue should build on initiatives that already provide important foundations rather than start from zero. These include the United Nations Educational, Scientific and Cultural Organization, UNESCO, Recommendation on the Ethics of Artificial Intelligence, which offers a global normative framework centered on human rights, transparency, accountability, and environmental sustainability. It should also connect with the Organization for Economic Cooperation and Development, OECD, AI Principles and the OECD AI Policy Observatory, which provide widely used policy guidance and comparative evidence across countries. The Global Partnership on Artificial Intelligence, GPAI, is also important because it links policy and expert collaboration on practical AI governance challenges. The International Telecommunication Union, ITU, AI for Good platform is another useful foundation because it connects technical communities, standards discussions, and development oriented applications of AI. The Council of Europe Framework Convention on Artificial Intelligence is also relevant as the first legally binding international treaty in this space. For my area of concern, the Dialogue should also connect with United Nations Water, because AI governance discussions still pay too little attention to water scarcity, river basin limits, and the human implications of expanding AI infrastructure. Data centre water use is not only a technical efficiency issue. It is also an issue of environmental governance, public accountability, and fair allocation of finite water resources. The added value of the AI Dialogue is that it can connect these efforts within one inclusive United Nations process, close the gap between digital governance and environmental governance, and elevate underdeveloped issues such as infrastructure transparency, water and energy reporting, and cumulative local impacts of AI growth.

Different stakeholders should contribute in distinct but connected ways. Governments can identify regulatory priorities and national needs. Industry can provide operational data, implementation lessons, and transparency on infrastructure impacts. Civil society can highlight rights, equity, and public accountability concerns. Researchers can contribute independent evidence, methods, and risk assessment tools. Local communities can ground the Dialogue in lived impacts, especially where AI infrastructure affects water, energy, land, and public services. The format should be designed to produce usable outcomes rather than a sequence of general statements. A strong structure would combine plenary sessions with smaller thematic working groups, each asked to produce a short list of practical recommendations. Each working group should include balanced representation from governments, technical experts, industry, civil society, and affected communities. The Dialogue should also require short evidence briefs in advance so discussion is informed by concrete cases rather than abstract positions. Sessions should include regional perspectives, since AI governance challenges vary across places and levels of capacity. For example, the implications of data centre expansion depend heavily on local water availability, drought risk, and infrastructure conditions. Finally, the Dialogue should include a public synthesis process that clearly records areas of agreement, unresolved tensions, and priorities for future cooperation. That would make participation more meaningful and would help ensure that contributions from different stakeholders shape real governance outcomes.

Several voices remain underrepresented in global discussions on AI governance. These include communities living near data centres and other digital infrastructure, water and energy sector experts, local and regional regulators, researchers from the Global South, Indigenous communities, workers in infrastructure and supply chains, and public interest groups focused on environmental justice and access to basic services. These perspectives matter because AI governance is often framed mainly around models, safety, and digital rights, while the physical footprint of AI receives less attention. Yet AI systems depend on land, electricity, cooling, and water in specific places. Communities experiencing drought, water stress, weak infrastructure, or cumulative industrial pressure may face real consequences even when they are far from the main centers of AI policy debate. Inclusion should be intentional, not symbolic. The Dialogue should reserve speaking and drafting roles for affected communities and sector experts, not only large institutions. It should provide financial support, remote participation options, and multilingual access so participation is not limited by resources or language. It should also invite evidence from local case studies, utilities, watershed managers, and community organizations. Meaningful inclusion requires recognizing that AI governance is not only about frontier technology. It is also about who bears the environmental and social costs of AI infrastructure, who has access to information, and who has a voice in shaping decisions that affect shared resources.

The most effective engagement formats will be those that move beyond formal speeches and create structured exchanges around real governance problems. One useful format would be case based policy labs where participants work through concrete examples such as data centre expansion in water stressed basins, cross border AI regulation, or public sector deployment risks. This makes discussion more practical and exposes tradeoffs clearly. Another strong format would be short evidence sessions that pair researchers, communities, regulators, and industry around one issue. For example, a session on AI infrastructure and water could include hydrologists, utilities, local communities, and data centre operators. That would help connect technical evidence with governance needs and lived experience. The Dialogue should also use facilitated working groups with a clear output, such as draft principles, reporting priorities, or recommended areas for cooperation. Open discussion is useful, but it should lead to written outcomes that can be reviewed and improved over time. To broaden participation, the process should include virtual regional consultations before the main event, multilingual submissions, and a public digital platform where evidence briefs and summaries are shared openly. This would allow more voices to shape the agenda before in person discussions begin. A meaningful Dialogue should be interactive, evidence based, and outcome oriented. The goal should not be only to speak, but to build shared understanding and practical pathways for action.

Several existing approaches offer useful models for effective AI governance at the global level. The United Nations Educational, Scientific and Cultural Organization Recommendation on the Ethics of Artificial Intelligence is a strong example because it connects AI governance to human rights, transparency, accountability, and environmental sustainability. The Organization for Economic Cooperation and Development AI Principles are also important because they provide a widely used framework for trustworthy AI that governments and organizations can adapt across different contexts. The European Union Artificial Intelligence Act shows the value of a risk based legal approach, while the United States National Institute of Standards and Technology Artificial Intelligence Risk Management Framework offers a practical model for identifying, assessing, and managing AI risks across the lifecycle of a system. International Organization for Standardization and International Electrotechnical Commission 42001 is also valuable because it provides a management system standard that helps organizations embed governance, accountability, and continuous improvement into AI practice. For the area I work on, a particularly useful practice is standardized environmental reporting. The Global Reporting Initiative 303 Water and Effluents Standard is a strong example because it requires organizations to report water related impacts in a more structured way. A similar approach should be encouraged for AI infrastructure, especially data centres, through disclosure of water withdrawals, water consumption, source water, timing of demand, and basin conditions. This would move governance beyond broad principles and toward measurable accountability. At a global level, one of the most promising approaches is to connect AI governance with infrastructure transparency, basin scale water assessment, and community level oversight so that AI development does not create hidden environmental costs across different regions and water stressed settings.