Independent Researcher in Public Law & Governance

Success for the first Global Dialogue on AI Governance would require three concrete outcomes. First, the adoption of a globally inclusive governance framework that explicitly recognizes the institutional absorption capacity of developing nations — not merely their technical readiness. A framework calibrated only for high-income contexts will fail at implementation in the Global South regardless of its technical soundness. Second, the establishment of binding environmental accountability standards for AI systems deployed in public services. Trustworthy AI cannot be environmentally extractive. Nations bearing the ecological costs of large AI models must have enforceable recourse within the governance architecture. Third, the creation of a Global South practitioner network — connecting mid-level public administrators, legal scholars, and governance researchers from Africa, Asia, and Latin America — to generate ground-level evidence that informs the next Dialogue cycle. Policy divorced from administrative reality produces legitimacy deficits, not governance improvements.

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

These four priorities reflect the core argument of my submission: that AI governance must be institutionally intelligent, not merely technically compliant. Safe and trustworthy AI requires transparency of decision chains within public administrations, not only algorithmic transparency. Human rights protection must function as a design constraint embedded before AI output generation, particularly in citizen-facing services where legal literacy is limited. Human oversight is meaningless without trained intermediaries - what I term Manager-Translators - capable of bridging AI outputs and administrative-legal realities. Finally, the social, cultural, and linguistic dimensions of AI are inseparable from governance legitimacy: systems that ignore local administrative cultures will generate institutional breaking points regardless of their technical performance.

Two critical cross-cutting issues remain insufficiently addressed by the current thematic framework. The first concerns energy consumption and environmental sustainability of data centers. The exponential growth of AI infrastructure - particularly large language models and cloud computing facilities - generates massive carbon and water footprints that disproportionately burden developing regions. Nations in Africa and the Global South bear the ecological consequences of AI infrastructure without equitable access to its benefits. The Dialogue must establish binding disclosure and mitigation standards for data center energy consumption as a prerequisite for trustworthy and equitable AI governance. The second concerns the legal and regulatory gap in AI legislation. Most nations - particularly in the Global South - lack adequate domestic legal frameworks to govern AI deployment, liability, and accountability. In the absence of enforceable national legislation, global governance standards remain aspirational. The Dialogue should prioritize a dedicated stream on AI legislative capacity-building, supporting states in developing context-sensitive legal frameworks that complement international standards rather than simply transplanting Western regulatory models. These two dimensions - environmental accountability and legal sovereignty - are foundational to any governance architecture that aspires to be genuinely global.

In Morocco and across North Africa, governance gaps in AI regulation create compounding challenges at three levels. At the institutional level, public administrations lack the legal frameworks and human capacity to evaluate, procure, or oversee AI systems responsibly. When AI tools are introduced into administrative processes — permit management, social services, land registration — without regulatory guardrails, they risk amplifying existing dysfunctions rather than correcting them. Administrators caught between AI-generated outputs and legally binding discretionary obligations face what I term Discretionary Paralysis: a systemic inability to act that erodes both efficiency and citizen trust. At the environmental level, data center expansion driven by AI demand increasingly strains energy and water resources in water-scarce regions like North Africa. Morocco's ambitious digital transformation agenda cannot be reconciled with climate commitments without binding AI sustainability standards at the international level. At the legislative level, the absence of a domestic AI legal framework leaves citizens without recourse when algorithmic decisions affect their rights. Courts lack the interpretive tools to adjudicate AI-related disputes, and public administrators lack the normative reference points needed for responsible AI integration. The opportunity, however, is significant. Morocco's position as a bridge between Africa, the Arab world, and Europe gives it a strategic role in shaping AI governance standards adapted to diverse administrative and legal cultures. The Global Dialogue represents a unique moment to ensure that this regional perspective shapes — rather than merely receives — international AI governance norms.

The AI Dialogue can play a transformative role in international cooperation by functioning as a legitimate translation space — bridging the gap between technically advanced governance frameworks produced in high-income contexts and the institutional realities of developing nations. Specifically, the Dialogue can add value in three ways. First, by establishing a common governance vocabulary that is legally operable across different administrative and legal traditions — not merely a Western regulatory export. Second, by creating peer learning mechanisms that allow mid-income and lower-income nations to share governance innovations adapted to resource-constrained environments, rather than simply receiving templates from above. Third, by institutionalizing the voice of frontline public administrators — those who actually implement AI-adjacent decisions at the local level — in the governance design process. Currently, AI governance is shaped predominantly by technologists, diplomats, and civil society organizations. The administrative practitioner perspective is structurally absent. The Dialogue's unique legitimacy as a UN-convened process gives it the authority to make these translations binding rather than merely aspirational.

The AI Dialogue should build strategically on three existing architectures. First, the UNESCO Recommendation on the Ethics of AI (2021) — already ratified by 193 member states — provides the broadest normative foundation. The Dialogue should operationalize its principles through concrete implementation indicators rather than producing parallel frameworks. Second, UNDP's governance capacity-building infrastructure across the Global South offers an existing delivery mechanism for translating governance standards into national legislative and administrative reform. The Dialogue should formally connect its outputs to UNDP country programming cycles. Third, regional bodies such as the African Union's AI continental strategy and ISESCO's digital governance initiatives represent context-sensitive frameworks that the Dialogue should integrate rather than bypass. The added value of the AI Dialogue lies precisely in its capacity to create coherence across these fragmented initiatives — functioning as a meta-governance layer that aligns existing mechanisms without replacing them.

Inclusive participation requires structural reform of the Dialogue's format. Three recommendations: First, establish dedicated tracks for administrative practitioners — not only governments, civil society, and technologists. Those who implement AI-adjacent decisions at the local level hold irreplaceable empirical knowledge that current formats systematically exclude. Second, create asynchronous written contribution mechanisms with multilingual support, so that participants from time-zone-disadvantaged or resource-constrained contexts can engage substantively without requiring real-time attendance. Third, publish a public contributors registry so that written submissions are formally attributed and citable — transforming the Dialogue from a consultation exercise into a genuine knowledge-production process.

Three communities remain structurally underrepresented. Mid-level public administrators from developing nations — those who translate policy into administrative action daily — are absent from global governance conversations despite being the primary implementation layer for any AI governance standard. Arabic, Swahili, and francophone African scholarly communities produce governance research that rarely enters international policy circuits due to language barriers and journal access inequalities. Rural and peri-urban citizen communities who interact with AI-adjacent administrative systems — without knowing it — have no channel to articulate their experience of algorithmic governance. Inclusion requires more than translation. It requires institutional investment in governance research capacity outside the traditional think-tank and university hubs of North America and Western Europe.

Three concrete approaches deserve attention. The EU AI Act (2024) represents the most advanced binding regulatory framework to date, establishing risk-based classification of AI systems with enforceable compliance obligations. Its value lies in operationalizing abstract principles into administrative procedures - a model adaptable to other legal traditions. Morocco's Digital Morocco 2030 strategy offers a practical example of a developing nation attempting to align digital transformation ambitions with governance capacity constraints. Its implementation challenges - particularly at the sub-national administrative level - provide valuable lessons for governance frameworks targeting the Global South. The Organizational Translation Model (OTM) - developed through field research in Moroccan administration - proposes a practitioner-centered approach to AI integration in public institutions. By training mid-level administrators as institutional translators between AI outputs and legal-administrative obligations, it addresses the human layer that most technical governance frameworks ignore. On environmental sustainability, recent research on the carbon and water footprint of large language models demonstrates that a single training run can consume millions of liters of water and emit hundreds of tons of CO₂ - costs disproportionately externalized onto water-scarce developing regions. Mandatory AI environmental impact disclosure standards, modeled on financial reporting requirements, represent a concrete and immediately actionable policy approach that the Dialogue should prioritize.