MBR School of Government, Center for Future of Government

The first Global Dialogue on AI Governance would be successful if it shifts global AI governance from general principles to practical cooperation that is meaningful for developing countries in the Global South, including the Arab region. 1- It should produce a clear, action-oriented agenda for closing AI divides: access to affordable compute, quality data, connectivity, skills, responsible AI tools, and institutional capacity. For the Global South, "safe and trustworthy AI" can only build on the capabilities to develop, procure, regulate, and audit AI systems. These capabilities are restricted or underdeveloped in most of the Global South. 2- It should create a structured pathway and mechanisms for underrepresented regions to influence global norms, with tangible commitments to interoperability. Many Global South countries currently operate as standard-takers, adapting to frameworks developed elsewhere. The Dialogue should produce a concrete pathway toward mutual recognition of standards, shared baselines for safe and trustworthy AI, and reduced compliance fragmentation. This is a key finding of our regional research. A successful inclusive Dialogue would enable transforming impacted stakeholders into standard-shapers through meaningful standing regional consultations, balanced expert representation, and mechanisms to feed national and regional priorities into the UN AI Scientific Panel and future UN processes. Designing effective and truly representative mechanisms for global governance would perhaps be the key outcome of the Dialogue. 3- It should identify a phased roadmap, with a small number of practical deliverables: For example, AI governance capacity-building packages; model regulatory guidance for resource-constrained governments; shared tools for risk assessment, impact assessment, audits, and public-sector procurement; and voluntary interoperability principles that do not impose one-size-fits-all models. 4- It should rely on building permanent multi-stakeholder follow-up architecture, built on AI-age conceptualization of inclusion. Governments, SMEs, academia, civil society, youth, language-specific experts (not just limited to official languages of the UN), women, people with disabilities, labor unions and communities affected by automated decision-making should have meaningful roles in decision making, not symbolic participation. Furthermore, universal data-representativeness should be a critical dimension to be prioritized. A successful Dialogue should also establish durable working groups that include SMEs (not only major technology firms), academia, civil society, and Global South governments, with clear deliverables and reporting cycles. 5- It should prescribe concrete steps on emerging AI frontiers. These include AI agents and autonomy, model and compute sovereignty, and equitable access to foundational infrastructure, among other areas where global rules remain underdeveloped and where Global South interests are particularly exposed. 6- It should help build trust by connecting AI governance to development priorities: public service delivery, education, health, climate resilience, food security, employment, and cultural and linguistic diversity. For the Global South, success means applying governance that protects rights while enabling innovation, regional competitiveness, and responsible public-sector transformation.

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

- Safe, secure and trustworthy AI: Safety and trustworthiness should be treated as a public value, a design principle as well as competitiveness priorities. In many frontier AI settings, they are seen as cost, distraction and compliance-related afterthoughts. In our regional research, cybersecurity, data privacy and security, explainability, and bias are seen as major developmental concerns. Governance should therefore support AI-ready cybersecurity baselines, privacy-preserving data sharing, AI incident reporting mechanisms, contestability channels, and sector-specific safeguards for high-stakes uses such as health, education, justice, employment, and public services. The Dialogue should focus on shared safety baselines that build trust without imposing prohibitive compliance costs on emerging economies. - AI capacity-building: Capacity-building should move beyond general training to include the "full stack" of AI readiness: digital infrastructure, affordable compute, data governance, regulatory capability, responsible AI expertise, procurement skills, R&D funding, and SME finance. Our research shows that infrastructure costs, internet reliability, limited risk capital, talent gaps, and weak IP protection constrain AI ecosystems in the Arab region, which is relevant to Global South contexts. Capacity-building should therefore support both governments, local innovators and societal groups. - Social, economic, ethical, cultural, linguistic and technical implications: AI governance must address distributional impacts. For the Global South, AI risks include widening economic dependency, job displacement without reskilling, cultural homogenization, and underrepresentation in datasets and models. For example, in the Arab region, Arabic-language AI, dialect diversity, cultural context, and locally relevant datasets are central governance concerns. Policies should support inclusive datasets, language technologies, and impact assessments that consider social and cultural effects. - Transparency, accountability and human oversight: Responsible AI remains difficult to operationalize. Governance of AI needs practical tools: audit trails, documentation standards, algorithmic impact assessments, independent review, explainability proportionate to risk, and clear human responsibility for AI-supported decisions. The goal should be accountable innovation, not compliance burdens that exclude and constrain.

Several cross-cutting issues require explicit attention. 1- AI sovereignty and dependency. Many Global South countries depend on foreign platforms, cloud infrastructure, foundation models, datasets, and standards. This creates risks of economic dependency, limited policy autonomy, and exposure to supply-chain or geopolitical disruptions. AI governance should address fair access to compute, data infrastructure, cloud options, and open, interoperable alternatives. 2- Public-sector AI governance deserves dedicated focus. Governments are major users and procurers of AI, yet many lack the capability to assess vendors, manage risk, protect data, and ensure accountability in automated or AI-assisted public services. Public procurement standards, model clauses, audit rights, and redress mechanisms should be developed for government use of AI. 3- AI agents and autonomous systems. There are continuously emerging system-level risks from AI agents, particularly where systems require broad access across devices and services. Governance must address user control, liability, security, and accountability for agentic AI. 4- Sustainable AI infrastructure should be recognized as a cross-cutting concern. Compute-intensive AI raises questions about energy, water, data-center sustainability, and equitable access to infrastructure. This is particularly relevant for regions balancing AI ambitions with climate and resource constraints. 5- AI market concentration. A small number of firms control large parts of the AI value chain. This affects competition, local innovation, data access, and bargaining power for SMEs. 6- Measurement and evidence in the field of AI are underdeveloped. Countries need comparable indicators on AI readiness, harms, adoption, capacity gaps, and governance effectiveness. The Dialogue should support shared metrics while allowing regional adaptation. 7- Labor market transitions and just transition. A dedicated focus on workforce transition, reskilling pathways, and protection of affected workers, particularly in economies with younger populations and elevated youth unemployment, is essential and only partially captured by "social and economic implications." 8- AI and algorithmic weaponization: Numerous powerful AI players are proactively seeking business models AI weaponizaion, not just in terms of utilizing AI in military activities, but also where AI is applied for societal control mechanisms, social engineering, political power abuse, countering participatory measures, among others. Large groups of the population in the Global South have been victimized already by these ongoing approaches. Addressing and controlling these growing powerful business models is a cross-cutting multilateral governance question.

For the Arab region, AI governance gaps create both strategic risks and major opportunities. The most significant challenge is that AI development is advancing faster than governance capacity. Many countries have national AI ambitions, but regulatory clarity, institutional coordination, audit capacity, and responsible AI implementation remain uneven. This uncertainty can slow innovation, raise compliance costs, and make it harder to scale across borders. Our research highlights that in the Arab region, regulation is experienced both as friction and as an enabler of trust, security, market access, and fairness. A second challenge is regional regulatory fragmentation. Divergent rules on data, privacy, cybersecurity, procurement, and AI accountability can limit cross-border AI services and reduce the ability to scale regionally. Third, capacity gaps affect the whole ecosystem. High electricity and internet costs, limited compute, uneven talent availability, insufficient venture financing, and weak IP protection constrain local innovation. Without addressing these enabling conditions, AI governance may become a compliance exercise rather than a driver of inclusive development. Fourth, social and cultural risks are significant. AI systems trained primarily on non-Arabic or non-regional data may fail to reflect Arabic language diversity, local norms, public-sector needs, and social contexts. This is a common challenge globally and can affect safety, fairness, accuracy, and public trust. The opportunity is equally clear: effective AI governance can become a competitiveness strategy and a driver for inclusive and sustainable development. Clear rules, responsible AI tools, capacity-building, regional harmonization, can help this region (like other Global South regions) move from AI adoption to AI creation, strengthen public services, and contribute more actively to global norms.

The AI Dialogue can play key core roles. 1- It can be a norm-setting forum with broad legitimacy. Unlike technical standards bodies or summit-driven processes, the Dialogue can convene under universal UN membership, granting it unique legitimacy to shape baseline norms on safety, transparency, and accountability that smaller economies can credibly endorse and adopt. This addresses a core finding from regional research: that many Global South countries are currently navigating frameworks "developed without their direct participation." 2- It can act as a bridge between global norms and national implementation. Many countries have developed high-level AI principles, but lack practical pathways to implement them. The Dialogue should translate principles into usable tools: model policy guidance, risk assessment templates, procurement standards, audit guidance, capacity-building programmes, and technical assistance options. 3- Act as a coordination mechanism for shared risks. AI-enabled misinformation, cyber threats, supply chain vulnerabilities, and frontier model risks transcend borders. The Dialogue can coordinate early warning, joint exercises, and collective response, complementing the work of national AI Safety Institutes. 4- It can become a coordination platform for capacity-building. The AI capacity divide is not only about skills; it includes compute, data, institutions, financing, regulatory expertise, and research capacity. The Dialogue can map capacity needs, connect countries with relevant UN agencies and partners, and reduce duplication among international initiatives. 5- It can support interoperability without imposing uniformity. Countries have different legal systems, development priorities, cultural contexts, and levels of readiness. The Dialogue should help identify common minimum expectations (human rights, safety, accountability, transparency, cybersecurity, and redress), while allowing flexible implementation. 6- For the Arab region and wider Global South, the Dialogue should also function as a voice-equalizing mechanism. It should ensure that developing countries are not merely consulted after standards are set, but involved early in defining priorities, evidence needs, and governance models. 7- The Dialogue can connect governance to the SDGs. AI cooperation should focus on public value: better education, health, climate resilience, food systems, public administration, and economic diversification. The UN resolution explicitly links the Dialogue to enabling AI to contribute to the SDGs and closing digital divides.

The AI Dialogue should connect and build on the existing landscape, rather than duplicate it. This includes building on: - UN system instruments. The UNESCO Recommendation on the Ethics of AI and its Readiness Assessment Methodology provide an established baseline already adopted by many member states. The UN Secretary-General's High-Level Advisory Body on AI and the Global Digital Compact provide complementary architecture. The ITU's AI for Good platform offers an operational track aligned with the SDGs. - Multilateral processes and partnerships. The OECD AI Principles, the Global Partnership on AI (GPAI), the Hiroshima Process Code of Conduct, and the Bletchley–Seoul–Paris-India AI Safety Summit series each contribute distinct elements that the Dialogue should integrate, such as principles, applied research, voluntary commitments, and frontier safety coordination. - Regional frameworks. The African Union's Continental AI Strategy, ASEAN's Guide on AI Governance and Ethics, the Council of Europe AI Convention, the EU AI Act's implementation experience, and Arab regional efforts through the GCC and Arab League digital initiatives offer regional governance experience worth synthesizing, even if at different maturity levels. - Technical and standards bodies. IEEE standards work, ISO/IEC JTC 1/SC 42, NIST's AI Risk Management Framework, and the network of AI Safety Institutes provide essential technical scaffolding. In reality, the Dialogue's distinct contribution is universal legitimacy, integrative function, and inclusive process. It can serve as a "framework of frameworks" — mapping interoperability, identifying gaps, structurally amplifying Global South voices that are underrepresented in the initiatives above, and providing a permanent home for issues no single body currently owns: agent governance, compute equity, and capacity-building at scale.

The Dialogue should invest in identifying the impacted stakeholders in each dimension of its work streams. This mapping will then lead the design of how different stakeholders should contribute according to their strengths. The following institutions are core: 1- Schools of Government and specialized centers of research (e.g. MBR School of Government and its Center of Future of Government) can intermediate with regional governments. These policy research institutions can provide independent evidence, regional analysis, evaluation frameworks, and capacity-building. Civil society can represent affected communities, rights concerns, inclusion, labor impacts, and access to redress. Technical experts can contribute practical tools for audits, assurance, cybersecurity, and evaluation. Through these institutions, governments can identify national priorities, regulatory gaps, public-sector use cases, and capacity needs. They can share lessons from AI strategies, data protection, cybersecurity, procurement, sandboxes, and public service delivery. 2- International organizations can provide technical assistance, comparative evidence, standards mapping, and capacity-building resources. They are also instrumental in connecting AI governance to SDG implementation. 3- SMEs and startups active in AI fields, can address local challenges practically, they illuminate how regulation affects innovation, compliance costs, investment, market access, and responsible deployment. Large technology companies and frontier model developers should contribute transparency on model capabilities, safety practices, data governance, and support for local ecosystems. In terms of format and organizational structures, the Dialogue should be: 1- Continuous, not episodic. Establish permanent thematic working groups operating year-round, with the annual plenary as a milestone rather than the entire process. 2- Hybrid by design. Combine in-person convening with robust digital participation to enable genuine Global South engagement and reduce travel-cost exclusion. Participation should not depend only on travel to Geneva or New York. Hybrid participation, regional hubs, multilingual submissions, and funded participation for Global South stakeholders are essential. The Dialogue should also include dedicated sessions for youth, SMEs, researchers, and affected communities. 3- Regional preparatory dialogues. Each cycle can begin with regional consultations (Arab, African, Latin American, Asian, Pacific) feeding common issues into the global process. 4- Issue-specific tracks. Dedicated streams on safety, capacity, accountability, agents, and inclusion, each with stable membership. 5- Transparent deliverables. Each cycle produces concrete instruments, such as toolkits, evaluation frameworks, not statements and policy recommendations alone. 6- Multi-language operation. Interpretation and document translation across UN languages, including Arabic, throughout the process.

Several voices remain underrepresented in global AI governance: - Global South governments. For example, despite being primary "rule-takers," the Arab region, sub-Saharan Africa, and parts of Latin America and Asia have limited representation in agenda-setting bodies. - Workers and labor representatives, particularly in sectors facing rapid AI-driven transformation, rarely have direct voice in governance fora. - Civil society from low- and middle-income countries, excluded by travel costs, accreditation barriers, and language requirements. - Women in AI from the Global South. Compounded gender and geographic underrepresentation persists across boards, technical roles, and policy panels. - Indigenous and minority communities, whose data and cultural knowledge are most at risk from extractive AI development, are systematically marginalized. - Speakers of underrepresented languages. Many languages remain underrepresented in foundation models, training data, and governance dialogue itself, producing both technical bias and discursive exclusion.

To move beyond traditional plenary speeches, the Dialogue can adopt the following formats: • Regional preparatory dialogues, regional pre-dialogue policy labs, with feed-through obligations. Each cycle begins with funded regional consultations producing inputs that the global process is required to address. These measures can create real accountability between regional and global levels. • Partnerships with Schools of Government on regional or local AI governance clinics. Countries or organizations could bring real policy challenges, such as public-sector AI procurement, audit requirements, data sharing, or sandbox design, and receive structured input from technical, legal, ethical, and development experts. • Joint scenario exercises and tabletop simulations. Multi-stakeholder exercises on cross-border AI incidents (cyber, misinformation, agent failures) build shared muscle memory and reveal governance gaps. • Standing thematic working groups with rotating chairs. Permanent groups on safety, capacity-building, accountability, agents, and inclusion, chaired in rotation across regions, generate continuity and distribute leadership. • Internationally connected policy sandboxes. National AI sandboxes (e.g., for high-risk uses, agentic systems, biometric applications) sharing methodologies, results, and lessons through a Dialogue-hosted repository — letting smaller economies learn without bearing full experimental cost. • Reverse mentoring and South–South–North learning. Structured exchanges where Global South policymakers brief Global North counterparts on regional realities (linguistic diversity, infrastructure constraints, SME needs), reversing the default flow of expertise. • Open digital deliberation platforms. Persistent online consultation hubs with multilingual interfaces, AI-assisted summarization, and transparent traceability from comment to text. • Policy-tech fellowships. Bringing technologists, policymakers, and civil society together to prototype governance tools — audit systems, transparency interfaces, explainability tools — with deliverables licensed openly. • Living libraries of case studies. Continuously updated, peer-reviewed repositories of governance experiments, evaluation methods, and incident learnings. • Citizen assemblies on AI. Demographically representative panels deliberating contested issues (facial recognition, generative AI in education) to surface public values, including from Global South contexts. • Youth and emerging-leader tracks. Dedicated channels for those who will live with AI's long-term implications. • Solutions marketplace showcasing practical tools: impact assessment templates, audit methods, procurement clauses, responsible AI training modules, open datasets, language-specific AI initiatives, and regulatory sandbox models. This would help make the Dialogue a platform for implementation, not only deliberation.

Capacity-building instruments • UNESCO SPARK-AI Alliance of AI-active Schools of public administration, and its MOOC on AI and Digital Transformation in Government • IEEE CertifAIEd AI Ethics Professional Training • UNESCO Readiness Assessment Methodology. Centers of Excellence • AI Safety Institutes. The growing network (UK, US, Singapore, Japan, EU, and others) demonstrates a model for distributed technical capacity that the Global South can replicate and connect into. • AI-focused Centers of Research-Practice-Policy (e.g. MBRSG Center for Future of Government - www.cfg.ae) Multi-stakeholder governance and transparency • GPAI working groups. Applied research that under-resourced governments can adapt to local needs. • OECD AI Policy Observatory. Sandbox approaches • Regulatory sandboxes in financial services provide controlled environments addressing the regulatory uncertainty widely reported by regional firms. PS: Submitted by Dr. Fadi Salem, Center for the Future of Government, Director, Policy Research Department, Mohammed Bin Rashid School of Government This submission is provided by the Mohammed Bin Rashid School of Government (MBRSG), a public policy research and teaching institution in the Arab world, through its Center for the Future of Government. The responses below are grounded in primary regional research conducted across 15 Arab countries, including input from hundreds of AI and digital companies complemented by high-level interviews, roundtables, and workshops with hundreds of leaders from the public and private sectors. The responses are written from a Global South perspective, with particular reference to the Arab region. They reflect a central conviction that emerged from the underlying research: safe, ethical, inclusive, clear, and interoperable AI governance is not a constraint on competitiveness. It is a competitiveness strategy. For much of the Global South, the AI divide is also a governance and capacity divide. The first UN Global Dialogue on AI Governance is therefore both an opportunity and a test: an opportunity to embed the perspectives of those currently positioned as standard-takers, and a test of whether multilateralism can deliver instruments, not merely declarations, that narrow the divide.