Open Transformation Lab Inc.

A successful first United Nations Global Dialogue on AI Governance should deliver three focused outcomes. First, it should create a practical framework for international cooperation on frontier AI risks while also addressing immediate harms already affecting societies today. Too much discussion focuses only on future existential scenarios, while many countries are dealing now with misinformation, fraud, bias, surveillance abuse, cyber threats, and disruption of jobs and public services. Success would mean recognizing both long-term and present-day risks within one shared agenda. Second, it should correct the imbalance in who shapes AI rules. At present, governance conversations are often dominated by a small number of powerful states and technology companies. The Dialogue should ensure meaningful participation from developing countries, especially those in Africa, Latin America, the Arab world, and South Asia, whose societies will be deeply affected by AI but whose voices are often underrepresented. A legitimate global framework cannot be built without them. Third, it should end with actionable commitments rather than symbolic statements. These could include support for regulatory capacity-building in lower-income countries, common transparency standards for advanced AI systems, channels for sharing safety research, and regular multistakeholder follow-up meetings under the United Nations system. Success would mean proving that AI governance can move beyond fragmentation and competition toward fair cooperation. If the first Dialogue establishes trust, inclusivity, and a concrete roadmap for next steps, it would mark an important turning point in global governance.

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

Our priorities reflect the need to ensure that AI governance responds not only to frontier technological risks, but also to the real and immediate challenges facing societies, particularly in the Global South. Safe, secure and trustworthy AI is essential because public trust is the foundation of any beneficial AI ecosystem. Systems that generate misinformation, enable fraud, reinforce repression, or operate unreliably can quickly damage confidence and create social harm. Safety must therefore include technical robustness as well as protection from misuse. Social, economic, ethical, cultural, linguistic and technical implications of AI are equally urgent because current AI systems often reflect structural inequalities. Many communities remain underrepresented in training data, languages outside dominant markets receive weaker performance, and local cultural contexts are frequently misunderstood. Without corrective action, AI can deepen exclusion rather than expand opportunity. AI capacity-building is a core priority because governance gaps are often capacity gaps. Many countries and institutions need support to develop regulatory expertise, technical literacy, independent research, auditing capacity, and public-interest innovation ecosystems. A fair global AI order requires broader participation, not concentration of knowledge and power. Transparency, accountability, and human oversight are necessary to make governance credible in practice. Citizens, regulators, journalists, and researchers need meaningful access to information about how systems are designed, deployed, and governed. Human oversight remains vital in high-impact areas such as media, healthcare, education, employment, and public administration, where automated decisions can carry serious consequences.

While the listed themes cover many core areas, several cross-cutting and emerging issues deserve clearer attention. First, the impact of AI on information integrity should be explicitly recognized. Generative AI is rapidly transforming the production, amplification, and personalization of misleading content, including disinformation, fraud, impersonation, and synthetic political messaging. This affects elections, public health, conflict environments, and trust in institutions. Second, concentration of power in the AI ecosystem requires greater focus. A small number of firms and states currently control advanced compute infrastructure, frontier models, cloud markets, and critical datasets. Governance discussions should address market concentration, dependency risks, and unequal access to technological benefits. Third, labor and knowledge transformation merit distinct treatment. AI is not only automating tasks; it is reshaping journalism, education, software development, creative industries, translation, and research. This raises urgent questions about fair transition policies, worker protections, authorship, and the future of expertise. Fifth, conflict and geopolitical uses of AI should be addressed more directly. AI is increasingly relevant to cyber operations, autonomous weapons, surveillance systems, sanctions evasion, and strategic rivalry. These uses can destabilize international security and widen inequalities between states.

In the Middle East and North Africa region, AI governance gaps are already having tangible effects across media, public discourse, and institutional development. Through my work with Arabi Facts Hub, which focuses on information integrity, media resilience, and AI impacts, we observe both serious risks and important opportunities. The most immediate challenge is the use of generative AI in misinformation ecosystems. Synthetic text, manipulated audio, automated propaganda, and scalable impersonation can intensify political polarization, reputational attacks, and public confusion—especially in fragile media environments. During crises or conflict, these tools can spread false narratives faster than institutions can respond. A second challenge is linguistic inequality. Arabic is one of the world's most widely spoken languages, yet many AI systems still underperform in Modern Standard Arabic and even more so in regional dialects. This creates quality gaps in translation, search, moderation, fact-checking, and public-service applications. It also risks importing cultural bias from models built primarily for other contexts. A third challenge is limited governance capacity. Many institutions in the region need stronger expertise in AI auditing, procurement standards, data governance, and regulatory design. Without this capacity, countries may become rule-takers rather than active shapers of AI norms. At the same time, the opportunities are substantial. AI can expand access to education, multilingual services, healthcare support, productivity tools, and more efficient public administration. In journalism, it can assist verification, translation, archiving, and investigative workflows when used responsibly. At Arabi Facts Hub, our response has been to train journalists and fact-checkers, study the use of large language models in Arab media environments, and build regional knowledge networks. The key opportunity for MENA is not simply adopting AI, but shaping trustworthy, inclusive, and Arabic-capable AI systems that serve public needs rather than deepen existing vulnerabilities.

The United Nations AI Dialogue can play a vital role by creating a trusted multilateral space where governments, researchers, civil society, and industry can coordinate on challenges that no country can solve alone. First, it can reduce fragmentation. AI governance is currently developing through separate national laws, regional frameworks, corporate standards, and ad hoc initiatives. While experimentation is valuable, excessive divergence may create regulatory gaps, duplication, and unequal protections. The Dialogue can help identify shared baseline principles—such as safety, transparency, accountability, and human rights protections—while respecting different national contexts. Second, it can elevate the voices of underrepresented regions. Many countries in the Global South, including across the Middle East and North Africa region, are affected by AI systems designed elsewhere yet remain underrepresented in rulemaking processes. The Dialogue can help ensure that multilingual societies, emerging economies, and smaller states participate meaningfully in shaping global norms. Third, it can mobilize practical cooperation. Beyond principles, countries need access to expertise, testing capacity, safety research, policy toolkits, and trusted knowledge-sharing channels. The Dialogue can support partnerships on capacity-building, regulatory training, and cross-border responses to harms such as AI-enabled fraud, disinformation, and cyber abuse. Fourth, it can build confidence between geopolitical rivals. AI is increasingly tied to strategic competition, but unmanaged rivalry could increase global risks. A neutral UN-led forum can encourage communication, transparency measures, and confidence-building steps even where broader political trust is limited.

The United Nations AI Dialogue should build on existing initiatives that have already demonstrated how cross-sector cooperation can produce practical governance ideas. One useful example is Columbia World Projects and its *Digital Governance for Democratic Renewal* initiative, developed with the Hertie School Centre for Digital Governance. That network convened researchers, regulators, policymakers, journalists, technologists, and civil society actors from both sides of the Atlantic to explore responses to platform power, transparency gaps, and democratic resilience. It also generated actionable recommendations on issues such as data access and accountability. The Dialogue should also connect with other ongoing processes, including the UNESCO Recommendation on the Ethics of AI, the OECD AI Principles, the G7 Hiroshima AI Process, regional frameworks such as the European Union AI Act, and civil society or academic networks working on auditing, safety, and inclusion. Its added value, however, should be distinct. First, unlike many existing forums, the AI Dialogue can offer universal legitimacy by bringing together developed and developing countries on equal footing. Second, it can connect fragmented efforts into a more coherent global ecosystem rather than duplicating them. Third, it can elevate underrepresented priorities—Arabic and African language inclusion, governance capacity-building, AI impacts on fragile information environments, and access for lower-income states. Fourth, it can convert principles into implementation through partnerships, technical assistance, and recurring review mechanisms. For actors such as Arabi Facts Hub working on AI and information integrity in the Middle East and North Africa region, the Dialogue would be most valuable if it becomes a bridge between global norm-setting and real local needs.

Different stakeholders should contribute according to their comparative strengths, while the Dialogue's structure should ensure balanced participation rather than dominance by states or major technology firms. Governments should share regulatory experiences, national strategies, procurement standards, and lessons from implementation. Private-sector actors should contribute technical expertise, transparency commitments, safety practices, and realistic assessments of deployment risks. Academic institutions and independent researchers can provide evidence-based analysis, benchmarking, foresight, and evaluation methods. Civil society organizations can represent affected communities, monitor rights impacts, and bring field experience on issues such as labor, inclusion, surveillance, and information integrity. Journalists and media organizations can help identify AI's effects on public discourse, misinformation, and democratic trust. International organizations can coordinate standards, technical assistance, and follow-up mechanisms. To be effective, the United Nations AI Dialogue should adopt a multi-layered structure. First, it should include a high-level plenary for political direction and endorsement of shared priorities. Second, it should organize focused thematic working groups on areas such as safety, capacity-building, transparency, human rights, and AI impacts on information ecosystems. These groups should produce practical recommendations rather than only general statements. Third, it should create regional consultation tracks so perspectives from the Middle East and North Africa region, Africa, Latin America, and Asia are systematically integrated rather than added symbolically. Fourth, participation should be hybrid and multilingual, with open calls for written submissions and remote engagement to include smaller institutions and experts with limited travel resources. Fifth, the process should include measurable follow-up: annual progress reviews, voluntary commitments, capacity-building partnerships, and publication of implementation reports.

Global discussions on AI governance remain more inclusive than before, but several important voices are still underrepresented—especially from the Arab world, Turkey, Iran, Pakistan, and Latin America. These regions represent hundreds of millions of people, major youth populations, diverse political systems, and fast-growing digital markets, yet their perspectives are often secondary to those of North America, Europe, and a small number of major technology powers. This underrepresentation matters because governance priorities differ across regions. In many Arab states and neighboring countries, key concerns include misinformation in conflict settings, language inclusion for Arabic and regional dialects, digital repression, youth employment, and fragile media ecosystems. In Turkey, Pakistan, Iran, and across Latin America, there are also urgent debates around platform power, data sovereignty, education, industrial upgrading, and democratic resilience. Global rules designed without these realities risk being incomplete or ineffective. Inclusion should move beyond symbolic invitations. First, the United Nations AI Dialogue should guarantee regional balance in panels, working groups, and expert advisory bodies. Second, multilingual participation is essential, including Arabic, Spanish, Turkish, Persian, and Urdu pathways for submissions and live interpretation. Third, travel support and hybrid participation should enable smaller institutions, researchers, journalists, and civil society groups to join meaningfully. Fourth, regional consultations should be held in the Middle East and North Africa and Latin America before global meetings, so agendas are shaped from the bottom up. Fifth, more partnerships with local universities, think tanks, and organizations such as Arabi Facts Hub can bring grounded expertise from affected communities.

Meaningful engagement depends on formats that generate substantive outputs. The most effective AI Dialogue would combine broad participation with smaller, task-oriented processes. First, structured written contributions should be central. Stakeholders should be invited to submit short policy memos, technical notes, case studies, and regional assessments in advance using a common template. These submissions could be synthesized into background papers before the Dialogue, ensuring that smaller institutions and experts who cannot travel still shape the agenda. A public repository of submissions would also improve transparency and cumulative learning. Second, closed working groups should be established to produce concrete outputs. These groups—balanced across governments, academia, civil society, and industry—could work under confidentiality rules that encourage candor and compromise. Each group should have a narrow mandate, timeline, and deliverable, such as recommendations on frontier model transparency, capacity-building for developing countries, synthetic media safeguards, or public-sector procurement standards. Their outputs should then be published for wider discussion. Third, "problem-solving labs" could be used instead of traditional panels. Participants would work through realistic scenarios: an AI-enabled election disinformation campaign, a cross-border fraud network, biased automated hiring systems, or misuse of biometric surveillance. This would move debate from abstract principles to operational cooperation. Fourth, regional roundtables should run in parallel, allowing regions such as the Middle East and North Africa, Latin America, Africa, and South Asia to identify shared priorities and feed them directly into the main process. Fifth, a commitments track could invite voluntary pledges from states, companies, universities, and NGOs, followed by annual reporting on progress.

A practical starting point would be mandatory transparency rules for advanced AI systems. Developers deploying high-impact models should disclose core information about training data sources, model capabilities, known limitations, energy use, safety testing, and major downstream partners. Regulators and accredited researchers should have deeper confidential access where public disclosure is not possible. Another urgent reform is a licensing or registration regime for frontier models above defined compute or capability thresholds. Systems with strong autonomous, cyber, persuasion, or synthetic media capabilities should face pre-deployment risk assessments, external audits, incident reporting duties, and recall or suspension powers if severe harms emerge. Governments should also require algorithmic impact assessments before AI is used in sensitive sectors such as healthcare, education, employment, policing, migration, welfare distribution, and credit. These assessments should examine discrimination risks, appeal mechanisms, human oversight, and proportionality, with public summaries available. To address the misinformation crisis, states and platforms should establish provenance standards for synthetic media. Watermarking, cryptographic content credentials, rapid labeling of AI-generated political advertising, and penalties for malicious impersonation would help protect information integrity while preserving legitimate expression. Competition policy is equally important. Antitrust authorities should scrutinize exclusive cloud arrangements, data monopolies, and acquisitions that entrench a few dominant firms. Public-interest compute infrastructure and shared research access for universities and smaller countries would reduce concentration. For developing regions, a global AI capacity fund should support regulators, universities, local-language datasets, testing labs, and technical training. Without this, governance will remain unequal and many countries will only consume rules written elsewhere.