Skidmore, Owings and Merrll

A meaningful outcome would be a shift in how intelligence itself is defined within governance frameworks. Today, most AI governance still assumes that intelligence is expressed through language and visual data. A successful Global Dialogue would expand this definition to include non-verbal, sensory forms of intelligence—touch, sound, smell, spatial awareness—recognizing that human experience and decision-making extend far beyond text and images. This matters because cities, buildings, and environments are not primarily understood through language. They are felt. If AI is increasingly embedded into the built environment, then governance must account for how systems interpret and act on sensory inputs that shape safety, comfort, and behavior. Without this, entire dimensions of human experience remain unrepresented—and therefore unprotected. Another outcome would be the introduction of "sensory accountability" as a governance principle. Just as we ask whether AI systems are fair or transparent, we should ask: what sensory realities are being amplified, ignored, or distorted? For example, how might an AI-managed urban system respond differently to noise patterns, air quality, or spatial density—and for whom? Finally, success would mean translating these ideas into operational standards. Not just principles, but testable requirements embedded into procurement, design workflows, and regulatory review. This is where organizations like United Nations Development Programme or UN-Habitat can play a critical role—bridging high-level frameworks with on-the-ground implementation in cities. The dialogue succeeds if it reframes AI governance not as managing algorithms, but as shaping the conditions through which humans sense, interpret, and inhabit the world.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI

My selection reflects a consistent line of thinking I have been developing across my work in architecture, writing, and AI governance: that intelligence in cities is not primarily linguistic or visual, but spatial and sensory. Most current AI governance frameworks are built around text, images, and data flows. Yet the environments where AI will have the most direct impact-cities, buildings, infrastructure-are experienced through movement, sound, air, light, and material conditions. This creates a structural gap. We are governing systems based on how machines "see," rather than how humans actually live. My focus on sensory inclusivity is an attempt to close that gap. It reframes AI not simply as a computational tool, but as a decision infrastructure embedded in physical space. In this context, governance must address how AI systems interpret and act on sensory inputs that influence safety, behavior, and well-being. For example, how an AI system responds to patterns of noise, crowding, or environmental stress is not just technical-it is ethical and spatial. This perspective is also shaped by my exposure to global policy discussions, including work connected to Economic and Social Council forums, where high-level principles often lack translation into spatial and operational realities. Institutions like UN-Habitat are beginning to bridge this, but the integration of sensory intelligence into governance remains underdeveloped. Ultimately, my selection is driven by a belief that the next phase of AI governance must move from abstract principles to embodied experience. If governance fails to account for how people actually sense and inhabit environments, it risks becoming precise in theory but incomplete in practice.

One emerging issue that remains underrepresented is the gap between AI as a technical system and AI as lived spatial experience. Current themes often address governance in terms of data, models, and regulation, but overlook how AI decisions materialize in physical environments-streets, buildings, and infrastructure where outcomes are directly felt. This creates a blind spot: systems may be compliant in code yet misaligned in reality. A second cross-cutting issue is the rise of AI as decision infrastructure across the real estate and urban value chain. From site selection and financing to design optimization and building operations, AI is no longer a tool at the edge-it is increasingly embedded at the core of decision-making. This introduces new governance challenges around accountability, liability, and risk transfer between designers, developers, operators, and capital providers. Organizations such as United Nations Development Programme and UN-Habitat have begun addressing urban data and governance, but the integration of AI into financial and operational decision layers is still evolving. Third, there is an emerging need to address non-visual, non-linguistic data regimes. As AI systems begin to incorporate inputs like soundscapes, environmental conditions, and movement patterns, governance frameworks must expand beyond traditional data categories. This connects directly to questions of inclusion: whose sensory realities are being measured, prioritized, or ignored? Finally, a critical gap lies in operationalization. Many discussions remain at the level of principles, while cities and organizations need actionable standards-procurement guidelines, certification systems, and audit mechanisms that translate governance into practice. Together, these issues point toward a broader shift: AI governance must evolve from regulating technologies to shaping the conditions under which people experience, trust, and live with those technologies.

In the U.S. built environment sector, governance gaps are emerging precisely where AI is moving fastest: into decisions that shape physical space, capital allocation, and daily experience. The most significant challenge is fragmentation. While technical guidance exists (e.g., risk frameworks and emerging regulations), there is no coherent bridge between digital AI governance and the realities of architecture, real estate, and construction. AI is already influencing site selection, underwriting, generative design, and building operations, yet responsibility remains diffuse. When an AI-informed decision leads to safety, compliance, or financial risk, it is often unclear whether accountability sits with the designer, developer, software provider, or operator. A second challenge is regulatory lag in spatial contexts. Existing governance structures are largely designed for data and software, not for environments that people inhabit. This becomes critical as AI systems begin to mediate conditions like density, airflow, noise, and circulation—factors that directly affect health and behavior but are rarely addressed in AI policy. Institutions such as National Institute of Standards and Technology are advancing risk management frameworks, yet translation into building codes, procurement standards, and professional liability structures remains limited. At the same time, the opportunities are substantial. There is a clear opening to define new standards at the intersection of AI and the built environment—from AI-informed design validation to operational governance in smart buildings and districts. Organizations like Urban Land Institute are beginning to surface these conversations within the real estate value chain, signaling market demand for clearer guidance. Another opportunity lies in positioning architects and designers as governance actors, not just users of AI tools. By framing AI as "decision infrastructure," the sector can lead in developing accountability models that integrate spatial, ethical, and financial considerations. Ultimately, the gap is not a lack of technology, but a lack of integration. The opportunity is to define that integration before it is imposed externally.

The AI Dialogue can play a pivotal role by shifting international cooperation from abstract alignment to operational interoperability. Today, regions are developing distinct governance models—risk-based regulation in Europe, standards-driven approaches in the U.S., and state-coordinated strategies elsewhere. The challenge is not the lack of frameworks, but their fragmentation. The Dialogue can act as a neutral space to translate these approaches into a shared "minimum layer" of governance: common definitions of risk, baseline accountability structures, and compatible audit methods that allow systems, firms, and cities to work across borders. A second role is to position AI as decision infrastructure within global systems, particularly in sectors like real estate, infrastructure, and urban development. Decisions about land use, capital flows, and building operations increasingly rely on AI, yet governance remains siloed by industry and geography. By convening actors from policy, finance, and the built environment, the Dialogue can surface where cross-border dependencies already exist—and where coordinated governance is necessary to manage systemic risk. Third, the Dialogue can advance inclusivity beyond language and vision by recognizing that human experience is spatial and sensory. International cooperation should not only align on data and models, but also on how AI-mediated environments affect safety, accessibility, and well-being. Organizations such as UN-Habitat and United Nations Development Programme are well positioned to connect these principles to real urban contexts. Finally, the Dialogue can accelerate implementation pathways—pilots, procurement guidelines, and certification systems that translate shared principles into practice. Cooperation succeeds not when frameworks converge in theory, but when they can be applied, tested, and trusted across borders. In this sense, the Dialogue becomes less a forum for agreement, and more a platform for building the infrastructure of trust between systems, institutions, and places.

A strong starting point is to connect existing efforts that each address a piece of the puzzle but rarely operate together. On the standards side, the National Institute of Standards and Technology AI Risk Management Framework provides a practical structure for identifying, measuring, and managing AI risk. In parallel, the European Commission's work on the EU AI Act advances a legally enforceable, risk-tiered approach. These establish technical and regulatory baselines, but they remain largely software- and data-centric. At the urban and development level, UN-Habitat and United Nations Development Programme have built extensive frameworks around sustainable cities, data governance, and inclusive innovation. Meanwhile, industry platforms such as the Urban Land Institute are beginning to engage real estate, capital markets, and operators in AI-related discussions, though without a unified governance structure. The gap is not the absence of initiatives, but their lack of integration across scales and sectors. The added value of the AI Dialogue lies in acting as a translation layer across these domains. It can align technical standards with urban policy, and connect regulatory intent with operational realities in sectors like real estate and infrastructure. This includes developing shared taxonomies for AI risk in physical environments, and linking them to procurement, certification, and investment decision-making. Another key contribution is advancing interoperability in practice—not only harmonizing principles, but enabling organizations to apply them consistently across jurisdictions. This could take the form of pilot projects, cross-border testing environments, or model governance protocols for AI embedded in cities. Finally, the Dialogue can introduce sensory and spatial dimensions into existing frameworks, ensuring that governance reflects how people actually experience AI-mediated environments—not just how systems process data. In essence, its value is not to replace existing initiatives, but to connect them into a coherent, actionable system.

Different stakeholders should contribute not as isolated voices, but as interdependent decision-makers within a shared system. Governments set regulatory direction and define accountability thresholds. Technology companies contribute technical architectures and risk visibility. Real estate, infrastructure, and finance actors determine where AI is actually deployed and scaled. Designers and architects translate these systems into lived environments, where decisions become tangible and consequential. International bodies such as UN-Habitat and United Nations Development Programme can anchor this collaboration by connecting global principles with local implementation. The Dialogue becomes effective when contributions are structured around shared decision points, not disciplines. For example: "AI in site selection," "AI in design approval," or "AI in building operations." Each stakeholder then contributes from their position within that decision chain, making gaps in responsibility and coordination visible. In terms of format, three layers would create both depth and execution: First, thematic roundtables that frame key governance challenges across sectors (risk, liability, data, and sensory impact). Second, scenario-based workshops where mixed stakeholder groups test how AI-driven decisions play out in real contexts—urban development, housing, or infrastructure systems. Third, implementation tracks that translate outcomes into actionable tools: procurement guidelines, audit frameworks, and pilot projects. To ensure continuity, the Dialogue should not end as a one-time convening. It should establish living working groups tied to specific sectors, producing iterative outputs and measurable progress. The added value of this structure is that it moves beyond discussion toward co-creation of governance in practice—where policy, technology, and spatial experience are shaped together, rather than in sequence.

A major gap in current AI governance discussions is the absence of spatial and sensory perspectives—the people who design, experience, and manage environments where AI decisions are actually felt. First, designers, architects, and urban practitioners remain underrepresented. AI governance is often framed by policymakers and technologists, yet many AI systems are already shaping buildings, infrastructure, and cities. Without these voices, governance risks overlooking how decisions translate into spatial consequences—safety, accessibility, and lived experience. Second, there is limited inclusion of non-verbal and sensory-based communities. People who navigate the world through sound, touch, or alternative sensory cues—such as individuals with visual or cognitive differences—offer critical insight into how environments function beyond language and imagery. Their perspectives are essential for understanding how AI systems may amplify or suppress certain experiences. Third, operators of the built environment—facility managers, maintenance teams, and local municipalities—are often missing from global conversations. These actors engage with AI systems in real time, yet governance frameworks rarely incorporate their operational knowledge. To include these perspectives, the Dialogue should move beyond traditional panel formats and adopt experience-based participation. This includes scenario workshops grounded in real environments (housing, transit, public space), where diverse users and practitioners can directly test AI-mediated conditions. Institutions such as UN-Habitat and United Nations Development Programme are well positioned to convene these voices at the intersection of policy and place, but the structure must intentionally prioritize them. Ultimately, inclusion is not only about representation, but about redefining what counts as knowledge in AI governance—from abstract data to embodied experience.

The AI Dialogue should move beyond conventional panels and use formats that make governance visible, testable, and embodied. One effective format would be scenario-based governance labs. Mixed groups of policymakers, technologists, designers, real estate actors, civil society, and affected communities could test AI decisions in realistic cases: housing allocation, infrastructure planning, smart building operations, disaster response, or public-space management. This would reveal where accountability breaks down in practice. A second format would be sensory and spatial simulations. Instead of discussing AI only through reports, participants could experience how AI-mediated environments affect movement, sound, accessibility, safety, and comfort. This would help include perspectives beyond language and visualization, especially from communities whose experiences are often underrepresented. Third, the Dialogue could host cross-sector "decision chain" workshops. Each session would follow one AI-enabled decision from data collection to model output, procurement, deployment, liability, and public impact. This structure would clarify who holds responsibility at each step. Finally, the Dialogue should include implementation studios. These would convert discussion into outputs such as procurement templates, risk-audit checklists, pilot proposals, and cross-border governance protocols. The most valuable format would combine reflection with production: not only asking what principles should guide AI, but building the first tools that allow those principles to operate across sectors, cities, and institutions.

Several existing approaches already point toward effective AI governance, especially when they move from principles to operational systems. At the standards level, the National Institute of Standards and Technology AI Risk Management Framework is a strong example. Its value lies in structuring AI governance around map, measure, manage, and govern, making risk an ongoing process rather than a one-time compliance check. Similarly, the European Commission's EU AI Act introduces a risk-tiered regulatory model, linking levels of risk to concrete obligations. This creates enforceability, though it remains largely focused on digital systems. At the organizational level, leading design and engineering firms such as Skidmore, Owings & Merrill have begun developing internal AI governance frameworks-including ethics guidelines, review processes, and structured adoption models. These are early but important steps toward embedding governance directly into professional workflows. On the urban side, UN-Habitat promotes people-centered smart city frameworks, emphasizing inclusion, sustainability, and local implementation. While not AI-specific, these provide a foundation for integrating governance into real environments. Another emerging practice is the use of digital twins-virtual models of buildings or cities that allow simulation of AI-driven decisions before deployment. When combined with governance protocols, they can serve as testing environments for safety, performance, and unintended consequences. The gap across all these examples is integration. Each operates at a different layer-technical, regulatory, organizational, or urban. The opportunity is to connect them into a continuous governance loop: standards inform regulation, regulation shapes procurement, procurement guides design and deployment, and real-world feedback updates the system. Effective AI governance will emerge not from a single framework, but from how these pieces are orchestrated into practice.