IramConsulting.ai

A successful Dialogue must begin with an honest recognition of what is being governed. We are no longer regulating software tools, but increasingly autonomous systems capable of decision-making, environmental interaction, and large-scale integration across education, public services, and economies. The speed of integration has outpaced institutional responsibility. First, academia must be recognized as a primary engine of governance capacity. Universities and colleges are scalable infrastructure for training policymakers, technologists, auditors, educators, and institutional leaders. Second, the Dialogue should establish baseline governance frameworks for agentic AI in institutional contexts, beginning with education. Agentic AI is already entering school systems. Children are among the most exposed and least protected stakeholders in this transition. Governance must include human-in-the-loop accountability, transparency of emergent behavior, and protection of cognitive, emotional, identity, and social autonomy. Third, governance must be contextual, not universal. One-size-fits-all models risk reinforcing global inequalities, particularly for developing nations and high-population emerging economies. Fourth, environmental accountability must be embedded directly into governance clusters, not treated as an afterthought. Inference-time compute is accelerating energy, water, and resource consumption, requiring coordinated global measurement and constraint. Fifth, data sovereignty must be treated as a first-order issue. When data generated in one country is owned and monetized elsewhere, especially by corporations based in more powerful economies, AI risks creating new forms of structural dependency. Finally, trustworthy AI cannot be achieved while monopoly structures continue to absorb innovation. The acquisition of emerging startups by dominant players is a governance risk, not merely a market dynamic. Success will require practical, scalable governance structures that align technological advancement with human capacity, institutional responsibility, data sovereignty, and planetary limits.

• Safe, secure and trustworthy AI
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight

These priorities are interconnected because we are no longer governing software alone. We are governing increasingly autonomous systems that will shape economies, institutions, learning, and human development. First, AI capacity-building is the foundation for all other governance. Global governance cannot be meaningful while many nations lack the institutional capability to participate, regulate, audit, and implement AI responsibly. Academia is one of the only scalable infrastructures capable of closing this gap by training policymakers, technologists, educators, auditors, and citizens. Second, safe, secure, and trustworthy AI must be redefined. Current frameworks often focus on technical and systemic risks while underestimating human impact. A more complete approach requires a human-layer understanding of safety that protects intellectual, identity, emotional, and social autonomy. This is especially urgent where agentic AI is entering schools and children are interacting with systems they may not fully understand. Third, the social, economic, ethical, cultural, linguistic, environmental, and technical implications of AI define the true scale of disruption. Governance must address cultural relevance, digital equity, environmental footprint, and data sovereignty. When data generated in one country is owned and monetized by corporations based elsewhere, developing nations face a new form of structural dependency. Finally, transparency, accountability, and human oversight are the operational mechanisms required for governing autonomous systems. As AI gains the capacity to reason, plan, and act, governance must mandate clear lines of responsibility, human-in-the-loop accountability, and transparency around emergent behavior. These priorities must also be protected from concentrated commercial influence. Trustworthy AI cannot be achieved while monopoly structures continue to absorb innovation and shape the direction of governance itself.

Yes. Several critical cross-cutting issues are insufficiently captured and will significantly shape the governance agenda. The first is environmental sustainability. Current frameworks do not adequately account for the planetary cost of AI infrastructure. The shift toward inference-time compute, where systems operate continuously rather than episodically, is accelerating energy, water, and resource consumption. AI's environmental footprint must be embedded directly into governance clusters, with coordinated international measurement, reporting, and constraint. The second is data sovereignty. For developing nations, this is a first-order governance issue. When data generated in one country is extracted, owned, and monetized by corporations based in another, AI governance risks reproducing structural dependency. This is not a future risk. It is already happening. The third is the governance of monopoly structures. The pattern of dominant players absorbing emerging AI startups is not only a market dynamic. It is a governance risk because it concentrates power over infrastructure, models, data, standards, and policy influence. The fourth is the governance of emergent behavior and value alignment in advanced AI systems. As frontier and agentic models demonstrate increasingly autonomous reasoning, planning, and adaptation capabilities, governance must extend beyond performance and safety to address how these systems develop and operationalize values over time. The fifth is the protection of children and young people. Agentic AI is already entering school systems, yet children are among the most exposed and least protected stakeholders. Addressing these issues requires a shift from reactive regulation to proactive governance. We must anticipate not only how AI is used, but how it evolves, who controls it, whose data powers it, and what planetary and social costs it creates.

Education is the sector where AI governance gaps are most consequential and urgent. The core failure is the continued treatment of AI as a tool, rather than as a structural intervention into human development, safety, and economic participation. The risks are immediate. Agentic AI is already entering school systems, and children are among the most exposed and least protected stakeholders. Young users may form emotional dependencies on AI systems designed to simulate human interaction, often without full awareness of their artificial nature. Cognitive offloading through daily reliance on AI tools may also affect critical thinking, decision-making, and social development. These effects are subtle, cumulative, and largely ungoverned. In parallel, the pathway from education to employment is being disrupted. Entry-level roles are being automated faster than institutions can adapt, leaving graduates underprepared for a workforce increasingly shaped by autonomous systems. Bias embedded in AI systems further risks scaling inequities across student populations in ways that may become difficult to reverse. Yet the opportunity is significant. AI enables personalized learning, scalable mentorship, and new pathways into entrepreneurship. In my own work, over 1,200 students have benefited from AI-augmented learning and venture development, demonstrating how individuals can create economic opportunity as traditional pathways change. For this reason, I have developed the Agentic AI Governance Framework for Schools, an original framework designed to protect minors in school-based AI deployment. It introduces an educational risk floor, human-in-the-loop accountability, developmental harm assessment, technical controls, equity monitoring, and minor consent architecture. The challenge is not whether AI should be used in education, but how it is governed. Without institutional-level agentic AI governance, these systems will shape learning, development, and opportunity invisibly and at scale.

The AI Dialogue can play a critical role by shifting international cooperation from discussion to structured alignment. At present, global AI governance remains fragmented, shaped by uneven regulatory capacity, competing national interests, concentrated market power, and rapidly advancing technological capabilities. The Dialogue provides an opportunity to establish a shared foundation that enables coordination without requiring uniformity. First, it can define common baseline principles while allowing for contextual implementation. Effective cooperation does not require identical policies, but it does require alignment on minimum standards, particularly in safety, accountability, transparency, data sovereignty, and human oversight for increasingly autonomous systems. Second, the Dialogue can strengthen global governance capacity. Significant disparities remain between countries in their ability to regulate, audit, and implement AI systems. Academia can serve as the backbone of this effort by training policymakers, technologists, educators, and auditors in governance literacy at scale. Third, it can facilitate coordinated approaches to cross-border challenges, including environmental impact, data governance, digital equity, cultural relevance, and the behavior of globally deployed AI systems. These issues cannot be addressed within national boundaries alone. Fourth, the Dialogue can help protect governance from concentrated commercial influence. Trustworthy AI cannot be achieved while monopoly structures continue to absorb innovation and shape the direction of the field. International cooperation must ensure that governance frameworks are shaped by collective human priorities rather than disproportionately influenced by a small number of dominant actors. The value of the Dialogue will lie not in consensus alone, but in its ability to create durable structures for coordination, accountability, capacity-building, and shared progress.

The AI Dialogue should build upon existing global and regional frameworks while addressing their structural limitations. Key initiatives include the UN AI Advisory Body's recommendations, the Hiroshima AI Process, the OECD AI Principles, UNESCO's Recommendation on the Ethics of AI, and the EU AI Act. Regional models such as Singapore's AI governance work also provide valuable operational guidance. However, these mechanisms share a common constraint. They largely approach AI as a system to be risk-managed, while AI is rapidly becoming increasingly autonomous, adaptive, and embedded across societal and institutional infrastructures. The AI Dialogue can add value in four ways. First, it can extend safety beyond technical risk to include a human-layer approach that protects intellectual, identity, emotional, and social autonomy. This is particularly critical in education, where children and young people interact directly with evolving systems. Second, it can introduce context-specific governance baselines. Agentic systems taking autonomous action in education, public service, healthcare, or critical infrastructure should default to higher-risk classification, with defined human-in-the-loop accountability. Third, it can operationalize capacity-building through academia. Higher education institutions can train not only users, but future policymakers, regulators, educators, and auditors. Fourth, it can elevate issues that remain underdeveloped in current frameworks, including environmental footprint, data sovereignty, digital equity, cultural relevance, and market concentration. The Dialogue can clarify that monopoly structures are not only economic concerns, but governance risks. By connecting with existing initiatives while expanding their scope to address autonomy, human impact, institutional readiness, sovereignty, and planetary limits, the Dialogue can move global governance from reactive compliance to proactive, capacity-driven leadership.

Effective contribution to the AI Dialogue requires moving beyond symbolic inclusion toward structured, role-based participation. Governments must provide regulatory direction and implementation lessons. Academia should serve as core governance infrastructure by contributing independent research, longitudinal evidence, and the training of future policymakers, auditors, educators, and technologists. Civil society and affected communities, including youth, educators, and populations in the Global South, must be integrated as co-designers, not observers. The private sector should contribute technical transparency, deployment data, and system-level insights under structured disclosure conditions. Environmental, infrastructure, and energy stakeholders must also be included because AI's energy, water, and resource footprint is now a governance issue. To support the Dialogue's work on safe, trustworthy, and rights-respecting AI, I have developed the Agentic AI Governance Framework for Schools, an original framework extending Singapore's Model AI Governance Framework for Agentic AI to school systems, with specific protections for minors. It introduces an educational risk floor: no agentic system acting on a child should be classified below Limited Risk, with High Risk as the default for any autonomous action affecting an individual student. The framework operationalizes this through four pillars: upfront risk bounding with developmental harm assessment; human-in-the-loop accountability with named sign-off; technical controls including prompt injection testing and disaggregated equity monitoring; and minor consent architecture for contexts where parental gatekeeping cannot be assumed. The Dialogue should establish sector-specific working groups, including education, healthcare, public services, data governance, and environmental sustainability. Outputs should include model governance frameworks, implementation toolkits, sector-specific risk thresholds, and standing review mechanisms. I would welcome the opportunity to share the full framework with the Co-Chairs and the relevant thematic discussion track ahead of the July 2026 session in Geneva. Framework Author: Iram Tanvir, AI Strategy & Foresight Advisor for Workforce Transformation; Professor; Founder, Iram Consulting.

The most consequential gap in global AI governance is not technical. It is representational. The voices shaping AI systems remain disproportionately concentrated among those with technical, financial, and geopolitical power, while those most affected remain underrepresented. Several groups are critically underrepresented. The Global South. Many developing nations are among the fastest-growing contexts for AI deployment, yet remain underrepresented in governance processes. Frameworks designed without their participation risk reinforcing structural dependency, especially when data generated in one country is owned and monetized by corporations based elsewhere. Children and young people. This generation will spend their educational and professional lives alongside increasingly autonomous systems, yet they have little formal voice in how these systems are governed. This is urgent because agentic AI is already entering school systems. Educators and academic institutions. Schools and universities are not only scalable infrastructure for AI capacity-building, but also the first line of defense against developmental, cognitive, and emotional risks. Communities experiencing algorithmic harm. Marginalized populations affected by biased systems in hiring, surveillance, public services, and education must be included as co-designers, not retrospective case studies. Environmental and infrastructure stakeholders. AI's energy, water, and resource demands are now governance issues. Experts in environmental systems, energy infrastructure, and sustainability must be included. Innovators and startup ecosystems also need representation. When emerging startups are repeatedly absorbed by dominant players, innovation diversity weakens and governance risks becoming shaped by concentrated market power. Inclusion requires structural change, not consultation. This means guaranteed representation for underrepresented regions, funded participation for lower-capacity nations, multilingual participation, youth and educator advisory mechanisms, and formal roles for civil society and interdisciplinary expertise in decision-making.

The AI Dialogue risks becoming what many international governance forums already are: a sequence of prepared statements that generate limited alignment or change. Avoiding this outcome requires structural innovation in how engagement is designed. Three principles should guide the format. First, consequence before consensus. Participants reason more effectively from concrete scenarios than from abstract principles. Engagement should begin with grounded cases across sectors such as education, healthcare, public services, data governance, and environmental sustainability, where governance failures have already occurred or are emerging. This shifts the discussion from "what should the rules be" to "what failed, who was affected, and who is accountable." Second, adversarial design over performance. For every proposed framework or principle, designated groups should be tasked with stress-testing it by identifying failure points, unintended consequences, populations left unprotected, environmental blind spots, and risks of regulatory capture. Governance systems strengthened through challenge are more durable than those built through consensus alone. Third, structural inclusion over invited participation. Asynchronous, multilingual engagement platforms should enable continuous input from underrepresented communities, particularly in lower-capacity nations. Meaningful inclusion cannot be confined to short sessions where access is shaped by geography, funding, language, or institutional status. The Dialogue should also include simulation-based governance labs. These could test scenarios involving agentic AI in schools, cross-border data extraction, autonomous decision-making in public services, market concentration, and AI infrastructure strain on energy and water systems. Finally, the Dialogue should establish a standing innovation mandate, commissioning institutions and practitioners with expertise in participatory foresight and governance simulation to continuously evolve engagement formats across sessions.

Existing governance frameworks provide important foundations, but many were designed to regulate AI systems as tools. They are not yet fully equipped to govern systems that are increasingly autonomous, adaptive, and embedded across institutional and societal infrastructure. Several examples offer meaningful and transferable value. The EU AI Act demonstrates how risk-based classification and enforceable obligations can be implemented at scale. UNESCO's Recommendation on the Ethics of AI establishes a global normative foundation grounded in human dignity, equity, sustainability, and human rights. The OECD AI Principles provide a valuable international reference point for responsible AI. Singapore's AI governance work offers operational guidance for accountability, transparency, and human oversight. Canada's Algorithmic Impact Assessment provides a practical model for embedding risk-proportionate governance within public sector systems. At the institutional level, embedding AI governance literacy, ethics, and applied use into higher education represents one of the most scalable approaches to building long-term governance capacity. Training future policymakers, technologists, educators, and auditors within academic systems ensures that governance capability is distributed and sustained. However, existing frameworks share a common limitation. They primarily define what AI systems must not do, while offering limited guidance on what they should be designed to achieve. As AI evolves toward more autonomous and agentic behavior, governance must expand beyond restriction toward direction, aligning system design with human development, institutional responsibility, cultural relevance, data sovereignty, and planetary limits. Effective approaches therefore share five characteristics: they are risk-proportionate, context-sensitive, institutionally embedded, environmentally accountable, and protected from concentrated commercial capture. The next phase of governance must advance from compliance-based models toward capacity-driven, purpose-oriented, and sovereignty-aware systems.