Independent

The first Global Dialogue would be a success if it produces outcomes that go beyond principles and translate into three concrete things. First, a shared risk taxonomy that governments, regulators, and practitioners can actually use. Most current frameworks use different terminology for the same concepts, making cross-border governance coordination unnecessarily difficult. The Dialogue should work toward a common language for AI risk that is specific enough to guide policy, not just aspirational. Second, a mechanism for bridging the implementation gap between global frameworks and local capacity. Countries in the Global South, including those in the MENA region, are adopting AI rapidly without the regulatory infrastructure or technical workforce to govern it responsibly. The Dialogue should not only produce frameworks but establish pathways through capacity-building, knowledge transfer, and regional support, for countries at different stages of readiness to implement them. Third, genuine multi-stakeholder input that shapes the final outputs. Written submissions from the Technical Community, civil society, and practitioners outside Western research institutions should visibly influence the Dialogue's conclusions. If the final documents reflect only the perspectives of governments and large technology organizations, the process will have produced legitimacy without representation.

• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Transparency, accountability, and human oversight
• Interoperability of governance approaches

These four priorities reflect what I observe as an independent AI governance practitioner and researcher working in Saudi Arabia and the broader GCC region. AI capacity-building is the most urgent priority for emerging markets. The gap between where global governance frameworks assume organizations are, and where they actually are in practice, is substantial. Without deliberate capacity investment, governance frameworks will exist on paper but not in practice. The social, cultural, linguistic, and technical implications of AI are underweighted in current global discussions. Most governance frameworks are developed in English by Western institutions. AI systems built primarily on English-language data perform differently across languages, including Arabic, creating not just quality disparities but governance blind spots. The Dialogue must address how frameworks apply across linguistic and cultural contexts, not only across legal jurisdictions. Interoperability of governance approaches is critical because organizations operating across multiple jurisdictions face a fragmented compliance landscape. The EU AI Act, NIST AI RMF, UNESCO Recommendation, and national frameworks like Saudi Arabia's use different terminologies and risk categories. A practitioner cannot simultaneously comply with all of them without significant duplication of effort. Common interfaces between frameworks would reduce this burden considerably. Transparency, accountability, and human oversight underpin all other governance priorities. Without them, frameworks become performative rather than functional. This is directly relevant to my contribution to a MIT-led study on AI risk prioritization, which identified accountability gaps as among the most critical and least addressed risks in current AI deployment.

Two issues are not adequately captured by the listed themes. The first is the governance gap specific to non-English AI systems. I have built a library if thousands of documented enterprise AI deployments, and the first comprehensive AI Glossary in Arabic and English. This work has made visible a significant gap: AI governance literature, risk frameworks, and compliance tools are overwhelmingly produced in English. Arabic-speaking populations represent hundreds of millions of AI users, yet the governance infrastructure (including terminology, assessments, policy templates, regulatory guidance) does not exist at adequate scale in Arabic. This is not only a linguistic issue; it reflects a deeper pattern where governance frameworks assume a Western institutional context (and culture!) that does not transfer directly to MENA and other regional settings. The second is the practitioner implementation gap. Governance frameworks are typically developed by researchers, governments, and large organizations. But most AI deployment decisions are made by mid-level managers, L&D professionals, HR teams, and small-to-medium enterprises that lack the resources to interpret complex governance documents. The Dialogue should consider how its outputs are translated into practitioner-accessible tools (assessments, checklists, plain-language guidance) that the people actually making deployment decisions can use. Without this, even well-designed frameworks will not change on-the-ground behavior.

Saudi Arabia and the broader GCC region present a distinctive case in the global AI governance landscape: rapid, large-scale AI adoption driven by national strategy (e.g. Saudi Arabia's Vision 2030 and SDAIA's SAMAI initiative, as well as a significant sovereign investment in AI infrastructure) occurring ahead of the governance frameworks needed to manage it responsibly. The most significant challenge is the implementation gap. Global frameworks such as the EU AI Act, the NIST AI RMF, and the UNESCO Recommendation exist, but they were designed for contexts with mature regulatory institutions, established AI auditing ecosystems, and large pools of governance-literate practitioners. Many organizations in MENA and GCC are deploying AI without adequate internal policy, risk assessment capability, or workforce literacy to govern those deployments responsibly. The frameworks exist globally; the capacity to apply them locally does not yet match the pace of adoption. The opportunity is significant. Saudi Arabia has the political will, the institutional infrastructure, and the financial capacity to become a genuine leader in AI governance for the Global South. What is needed is a bridge between global frameworks and local implementation: practitioner-level tools, Arabic-language governance resources, and capacity-building programs grounded in the specific regulatory and cultural context of the region. The Global Dialogue is well positioned to support exactly this.

The AI Dialogue occupies a position no other body can: a universal, intergovernmental space convened under the authority of the General Assembly, where both major AI-producing nations and countries that are primarily AI-adopters have equal standing. That universality is its most significant and irreplaceable asset. The role it can play most distinctively is as a bridge between the governance frameworks being produced primarily in the Global North and the realities of AI adoption in the Global South, where the pace of deployment often outstrips regulatory capacity, and where the frameworks themselves frequently do not account for local institutional, linguistic, and cultural contexts.

The Dialogue should build on what already exists (e.g. OECD AI Principles, UNESCO's Ethics Recommendation, and GPAI's technical work) rather than duplicating them. At the regional level, it should study how frameworks like the EU AI Act and Saudi Arabia's National AI Strategy are actually being implemented, learning from both successes and friction points. Its unique added value is UN-level universality. The Dialogue speaks for everyone, and that legitimacy should be used to harmonize existing frameworks and establish minimum shared standards that practitioners across all jurisdictions can actually work with.

The AI Dialogue will be most effective if it moves beyond a format in which different stakeholders simply present their positions sequentially and instead structures contributions by function, not just by category. Governments bring regulatory authority and implementation experience. Their most valuable contribution is not statements of intent but concrete reporting on what governance measures are working, what is failing, and what resources they need. The Dialogue should ask governments to report on implementation realities, not just policy positions. The Technical Community and independent practitioners bring implementation-level evidence that is systematically absent from most intergovernmental AI discussions. Their contributions should surface concrete cases, where governance frameworks have been applied, where they have failed, and what practitioner-level tools are missing. Written submissions alone are insufficient; dedicated working sessions with practitioners should feed directly into the Dialogue's outputs. Civil society brings the perspective of affected communities: those experiencing AI's impacts in healthcare, employment, education, and public services without meaningful input into how those systems are governed. Their contributions should shape the substance of outputs, not merely appear in the participant list. On format, the Geneva sessions in July 2026 should not be the starting point. Regional consultations, feeding structured inputs from different geographies into a synthesis document, should precede the main sessions. This would ensure that the Geneva Dialogue reflects genuinely global input rather than the perspectives of those with the resources to participate in person.

Several communities are systematically underrepresented in global AI governance discussions, and their absence has direct consequences for the quality of frameworks produced. Independent practitioners (consultants, trainers, and advisors who work directly with organizations deploying AI) occupy a critical position in the implementation chain but have no formal voice in governance processes. They are neither large technology companies nor academic researchers, yet they are often the people translating governance frameworks into organizational practice. Their ground-level experience of what works and what does not is largely invisible to the bodies producing frameworks. Arabic-speaking communities and professionals represent hundreds of millions of AI users whose primary language is almost entirely absent from governance discourse. Governance frameworks, risk taxonomies, compliance tools, and training resources are overwhelmingly in English. Including these communities requires not just translation but active consultation in Arabic, and the production of governance resources accessible to Arabic-speaking practitioners and policymakers. Small and medium enterprises, which constitute the majority of businesses deploying AI in most economies, are absent from governance discussions dominated by large technology companies and national governments. Yet SMEs often lack the legal, compliance, and technical resources to navigate complex governance frameworks, and their deployment decisions affect millions of people. Finally, non-Western ethical frameworks for thinking about responsibility, harm, and fairness in technology have been largely absent from the philosophical foundations of AI governance. Including these perspectives would strengthen the universality that the Dialogue is designed to represent. Inclusion requires more than open submission portals. It requires active outreach, translation, regional partnerships, and dedicated sessions that bring these communities meaningfully into the process.

I think the most common format in intergovernmental AI discussions (panel presentations followed by general debate) produces declarations rather than decisions. Several alternative formats would generate more meaningful and durable engagement. One is structured scenario exercises, in which participants from different stakeholder groups work through a concrete AI deployment case and identify where current governance frameworks succeed or fail, would ground the Dialogue in operational reality. A healthcare AI system operating across three jurisdictions with different regulatory requirements reveals more about framework gaps than any abstract discussion of design principles. Also, red-teaming sessions, in which groups are specifically tasked with identifying weaknesses in proposed governance approaches, would strengthen the quality of outputs. Governance frameworks are typically tested against best-case assumptions; deliberate adversarial review surfaces the gaps that matter most in practice. Working groups that produce actionable outputs (like a draft risk taxonomy, a practitioner checklist, a model policy template) rather than only recommendations, would give the Dialogue something concrete to show for each session. Outputs that practitioners can use immediately demonstrate that governance discussions have real-world consequence. For participants unable to attend Geneva in person, structured asynchronous participation (not just written submissions, but facilitated online working groups with real influence on outputs) would meaningfully expand the diversity of voices without requiring travel. Finally, publishing working drafts and inviting structured public comment during the Dialogue process, not only through pre-session submissions, would maintain momentum and signal that the process is genuinely open to the communities it is designed to serve.

Several policies and practices offer concrete models worth highlighting. I see that the EU AI Act's risk-tiered approach, classifying AI systems by risk level and applying proportionate obligations accordingly, is the most structurally coherent regulatory model to date. Its value lies not in its specific rules but in the principle: governance requirements should scale with the potential for harm, rather than applying uniform compliance burdens to all AI systems regardless of context. The NIST AI Risk Management Framework offers a complementary practitioner-level model. Unlike the EU AI Act, it is not a compliance instrument but a voluntary framework for organizations to assess, manage, and document AI risk. Its profile-based approach allows organizations to adapt the framework to their specific context, making it more applicable across sectors and institutional sizes than prescriptive regulation. At the national level, Saudi Arabia's SDAIA has demonstrated how an apex AI authority can coordinate AI strategy, governance, and capacity-building under a single institutional mandate. For example the SAMAI initiative delivered AI literacy training to over one million citizens and is a concrete example of how governance and capability-building can be pursued simultaneously rather than sequentially. From a practitioner standpoint, organizational AI readiness assessments before deployment represent an underutilized but scalable approach. Publicly available assessment frameworks, deployed through training centers, L&D programs, and regulatory guidance, can extend governance reach to the SMEs and mid-size organizations that formal regulation rarely touches effectively. Finally, the principle of open, publicly accessible AI governance resources like taxonomies, risk registries, and policy templates lowers the barrier for smaller organizations and lower-income countries to participate in responsible AI adoption without requiring expensive external consultants.