MagdalenaGovernance.ai

A successful first Global Dialogue on AI Governance would, in my view, be measured less by lofty declarations and more by whether it creates durable alignment, practical next steps, and trust across very different stakeholders. First, success would mean establishing a shared baseline of principles: safety, human rights, accountability, transparency, fairness, and meaningful human oversight. The goal does not need to be full legal harmonization, but there should be enough convergence to reduce fragmentation and signal that AI governance is a common global responsibility. Second, the Dialogue should produce a clear roadmap for cooperation. That could include agreement on priority areas such as frontier AI safety, public-sector use of AI, election integrity, protection of children, and governance of general-purpose models. A strong outcome would be a commitment to continued working groups, timelines, and named owners, rather than a one-off conversation. Third, success depends on inclusive participation. The process should not be dominated only by major powers or large technology companies. It should meaningfully include the Global South, civil society, academia, technical experts, and affected communities. If countries with fewer resources leave feeling heard and supported, that would be a major achievement. Fourth, the Dialogue should encourage practical capacity-building: technical assistance, shared standards, policy toolkits, and support for regulators and public institutions that are still building AI expertise. Governance cannot succeed if only a few jurisdictions can implement it. Finally, success would mean building trust without freezing innovation. The best outcome is a credible message that innovation and governance are not opposites: good governance makes beneficial AI more legitimate, more secure, and more widely accepted. In short, the Dialogue will be successful if it turns broad concern into shared principles, ongoing cooperation, and actionable commitments.

• Safe, secure and trustworthy AI
• AI capacity-building
• Transparency, accountability, and human oversight
• Open-source software, open data and open AI models

I selected these priorities because they are the foundations of credible and inclusive AI governance. Safe, secure and trustworthy AI is essential because public trust in AI will depend on whether systems are reliable, resilient, and designed to prevent harm. Without safety and security, adoption will be fragile and the social and economic benefits of AI will be undermined. AI capacity-building is equally important because effective governance cannot be limited to a small number of technically advanced states or large companies. Many countries and public institutions need support to build regulatory, technical, and operational expertise. Capacity-building is what makes global AI governance more equitable and more realistic in practice. Transparency, accountability, and human oversight are necessary to ensure that AI remains subject to human responsibility. People affected by AI systems should be able to understand when AI is used, who is responsible for outcomes, and how decisions can be challenged or reviewed. These principles are central to legitimacy, rights protection, and democratic trust. I also selected open-source software, open data and open AI models because openness can support innovation, research, security testing, and broader access to AI development. Open ecosystems can help reduce dependency on a small number of dominant actors and can strengthen global participation. At the same time, openness should be accompanied by appropriate safeguards, especially where misuse or systemic risk is possible. Taken together, these four areas reflect a balanced approach: AI should be safe and trustworthy, governed responsibly, accessible beyond a few powerful actors, and supported by the knowledge and institutional capacity needed for meaningful implementation.

Yes - AI literacy for society is a major cross-cutting issue that should be treated as a distinct priority, not only as a sub-part of capacity-building. Public AI literacy matters because AI governance will not succeed if only policymakers, regulators, and developers understand how AI works. Citizens, workers, teachers, journalists, parents, and consumers increasingly interact with AI systems in everyday life. They need the ability to recognize when AI is being used, understand its opportunities and limitations, question outputs critically, and know when human review is needed. AI literacy is also closely linked to democratic resilience. In a world of synthetic content, deepfakes, automated persuasion, and AI-assisted misinformation, societies need stronger civic and digital literacy to protect public debate and trust. This is especially important for children and young people, who are among the most exposed to AI-enabled environments. In addition, AI literacy supports fairness and inclusion. Without broad public understanding, existing inequalities may deepen: some groups will benefit from AI tools and opportunities, while others may be excluded, manipulated, or disproportionately harmed. Literacy should therefore include not only technical awareness, but also understanding of rights, risks, bias, privacy, and available remedies. I would also highlight environmental sustainability as an emerging cross-cutting issue that deserves more visibility. The energy, water, and infrastructure demands of AI systems are becoming increasingly relevant to responsible governance. In short, AI literacy for society should be recognized as a foundational enabler of effective governance. Safe, transparent, and accountable AI requires not only good rules and capable institutions, but also an informed public able to engage with AI critically and confidently.

In both the European and African contexts, and particularly in sectors handling personal data, education, research, public services, and digital innovation, governance gaps in safe and trustworthy AI, transparency and accountability, capacity-building, and openness are already having visible effects. The most significant challenge is the gap between rapid technological deployment and institutional readiness. Many organizations and public bodies are adopting AI tools faster than they can assess legal, ethical, security, and operational risks. In Europe, this creates pressure around compliance, accountability, and implementation. In many African countries, the challenge is even sharper because regulatory frameworks, enforcement capacity, technical infrastructure, and access to expertise may still be developing. A second major challenge is uneven capacity. Better-resourced actors are more able to invest in governance, testing, procurement safeguards, and internal controls, while smaller institutions, startups, universities, and public authorities may struggle. In Africa, this can deepen dependency on external providers and widen digital inequalities. In Europe, it can create uneven implementation across sectors and institutions. At the same time, there are major opportunities. For Africa, AI can support development priorities, including education, agriculture, healthcare, financial inclusion, and public administration, provided deployment is context-sensitive and rights-respecting. For Europe, stronger governance can reinforce trustworthy innovation, legal certainty, and public confidence. In both regions, open-source and open AI models can support local innovation, collaboration, and technological sovereignty, if paired with appropriate safeguards. Overall, the key issue is not whether AI will be used, but whether governance evolves quickly enough to ensure that adoption is safe, inclusive, explainable, and beneficial. A major opportunity lies in stronger Africa-Europe cooperation on capacity-building, AI literacy, standards, and shared governance approaches.

The AI Dialogue can play a valuable role as a bridge-building forum: not to replace existing national, regional, or multilateral initiatives, but to connect them, reduce fragmentation, and create space for practical cooperation. First, it can help build a shared language and baseline understanding across countries with different legal systems, levels of technical development, and policy priorities. That is especially important in areas such as safety, accountability, transparency, human oversight, and the governance of general-purpose and open models. Second, the Dialogue can support greater policy coherence. Today, one of the main risks in AI governance is divergence: different rules, standards, and expectations developing in isolation. The Dialogue can encourage interoperability of approaches, mutual learning, and the identification of minimum common principles, even where full harmonization is neither possible nor desirable. Third, it can advance capacity-building as a form of cooperation. International cooperation on AI governance will only be credible if it includes support for countries and institutions that are still developing regulatory, technical, and enforcement capacity. The Dialogue can help mobilize training, technical assistance, policy toolkits, and partnerships, particularly for the Global South. Fourth, it can promote inclusive participation and trust. A meaningful dialogue should bring together governments, international organizations, academia, civil society, technical experts, and industry, while ensuring that Africa and other underrepresented regions are not merely participants, but agenda-shapers. Finally, the Dialogue can serve as a platform for action-oriented follow-up: working groups, joint statements, shared research, cooperation on standards, and mechanisms for ongoing exchange. In that sense, the AI Dialogue's greatest value lies in turning abstract concern into structured cooperation: helping countries move from parallel discussions toward more coordinated, inclusive, and effective AI governance.

The AI Dialogue should build on, and connect with, existing global and regional frameworks rather than duplicate them. Key foundations include the OECD AI Principles and the Global Partnership on AI (GPAI), which already provide a strong basis for human-centric, safe, secure, and trustworthy AI and for practical international cooperation. UNESCO's Recommendation on the Ethics of AI, adopted by all 194 Member States, is also critical because it offers a globally recognized normative framework grounded in human rights. The UN Global Digital Compact is directly relevant, as it provides the wider multilateral umbrella for digital cooperation and AI governance. In Europe and beyond, the Council of Europe Framework Convention on AI is important as the first international legally binding treaty in this field. Regional initiatives should also be connected, including the African Union Continental AI Strategy, which brings an Africa-centric, development-focused perspective. Practical and technical communities such as ITU's AI for Good can also contribute through standards, skills, and implementation partnerships. The added value of the AI Dialogue would be its ability to act as a convening and coordination platform across these initiatives. It can help translate parallel efforts into a more coherent global conversation, reduce fragmentation, and identify areas of interoperability rather than forcing uniformity. It can also fill an important gap by bringing together states, international organizations, technical experts, civil society, and underrepresented regions in one ongoing process. Most importantly, it can elevate issues that are often unevenly addressed across forums, such as capacity-building, AI literacy, inclusion of the Global South, and practical follow-up mechanisms. In that sense, the Dialogue's value is not in replacing existing work, but in connecting it, widening participation, and turning principles into sustained cooperation.

Different stakeholders should contribute to the AI Dialogue in complementary and clearly defined roles. Governments should bring policy experience, identify regulatory priorities, and share implementation challenges. International organizations can help connect existing frameworks, provide evidence, and support coordination across regions. Industry should contribute technical expertise, real-world deployment insights, and practical information on risks, safeguards, and innovation needs. Civil society should ensure that human rights, inclusion, consumer protection, and impacts on affected communities remain central. Academia and technical experts can provide independent research, foresight, and evaluation methods. Youth, educators, and public-interest actors should also be included, especially on AI literacy and societal impacts. To be effective, the Dialogue should be structured as a multi-stakeholder, action-oriented process, not a one-off event. A useful format would combine: High-level plenary sessions to identify shared priorities and political direction; Focused thematic working groups on issues such as safety, capacity-building, transparency, open models, and AI literacy; Regional and sectoral consultations to reflect different legal, economic, and social realities, including African perspectives and those of the Global South; Practical follow-up mechanisms, such as summary recommendations, voluntary commitments, joint capacity-building initiatives, and a roadmap for future meetings. The Dialogue should also be designed for balanced participation. Representation should not be limited to major powers or large technology companies. Smaller states, developing countries, SMEs, researchers, and affected communities should have meaningful speaking opportunities and agenda-shaping roles. Overall, the Dialogue will be most valuable if it is inclusive, transparent, and continuous: a forum where different stakeholders do not simply exchange views, but help shape practical cooperation, build trust, and support the responsible governance of AI across regions and sectors.

Several voices remain underrepresented in global AI governance discussions, even though they are often the most affected by AI systems. First, countries from the Global South, especially from Africa, parts of Latin America, and smaller developing states, are still too often invited as participants rather than treated as agenda-shapers. Their perspectives are essential because AI governance must reflect different development needs, infrastructure realities, linguistic contexts, and institutional capacities. Second, affected communities are often missing from the room: workers subject to algorithmic management, students, migrants, persons with disabilities, minority groups, children and young people, and communities exposed to surveillance or exclusionary public-sector systems. Governance discussions are weaker when they do not include those who experience AI's real-world impacts directly. Third, smaller innovators and public-interest actors are underrepresented. Debate is often dominated by governments, large technology firms, and a limited set of expert institutions. SMEs, open-source communities, independent researchers, educators, journalists, consumer groups, and grassroots civil society organizations should have a stronger role. Fourth, non-English and non-Western knowledge traditions remain insufficiently visible. AI governance should not rely only on a narrow set of legal, technical, or cultural assumptions. These groups can be included through more deliberate design of the Dialogue: balanced regional representation, funded participation for under-resourced actors, multilingual access, open consultations, youth and civil society seats in thematic working groups, and structured channels for affected communities to provide evidence and recommendations. Inclusion should also mean influence: not only attending sessions, but helping shape agendas, draft outcomes, and define priorities. In short, AI governance will lack legitimacy if it is discussed mainly by the most powerful states and companies. A credible dialogue must include those with fewer resources, different lived realities, and perspectives that are too often treated as peripheral, even though they are central.

To foster meaningful and dynamic engagement, the AI Dialogue should move beyond formal statements and use formats that encourage interaction, problem-solving, and inclusion. One effective format would be moderated roundtables with a balanced mix of governments, industry, civil society, academia, and underrepresented regions. Smaller groups often produce more candid and practical discussion than plenary sessions. A second useful format is scenario-based policy labs. Participants could work through realistic cross-border cases — for example, a high-risk public-sector AI system, an open-model misuse incident, or AI-generated election disinformation. This helps stakeholders test governance ideas against real-world complexity rather than speaking only in abstract terms. Lightning interventions could also improve dynamism. Short, tightly timed contributions from diverse participants — including youth, SMEs, researchers, and affected communities — would widen participation and prevent discussion from being dominated by a small number of voices. Another valuable tool would be regional and sector-specific breakout sessions, allowing participants to discuss governance needs in contexts such as Africa, education, healthcare, labour, or public administration. This would help ensure that the Dialogue reflects different realities, not only global generalities. The process could also include interactive consultation mechanisms: digital input platforms, live polling, and pre-submitted stakeholder questions. These can help surface areas of convergence and disagreement in real time. Finally, the Dialogue should end with co-creation sessions focused on drafting practical outputs: joint recommendations, voluntary commitments, capacity-building partnerships, or priorities for future working groups. In short, the most effective engagement formats will be those that are multi-stakeholder, participatory, and action-oriented — designed not only for exchange of views, but for building trust, testing ideas, and producing usable outcomes.

Examples of effective AI governance already exist at several levels: binding rules, international principles, management standards, and practical risk tools. A leading policy example is the EU AI Act, which applies a risk-based approach by prohibiting certain unacceptable uses, imposing requirements on high-risk systems, and introducing rules for general-purpose AI. Its significance is that it translates broad governance values into concrete legal obligations. At the international level, the OECD AI Principles and the UNESCO Recommendation on the Ethics of AI provide widely recognized foundations for trustworthy, human-centred AI governance. They are valuable because they promote common expectations around transparency, accountability, human rights, fairness, and international cooperation across different jurisdictions. UNESCO's instrument is especially important because it was adopted by all 194 Member States. For organizational implementation, the NIST AI Risk Management Framework and ISO/IEC 42001 offer concrete solutions. NIST provides a practical structure for identifying, assessing, and managing AI risks across the lifecycle, while ISO/IEC 42001 helps organizations build internal AI governance systems, including accountability, continual improvement, and risk controls. A useful additional example is the MIT AI Risk Repository, which helps map and classify AI risks in a structured way. Tools like this are important because effective governance requires not only principles and laws, but also shared taxonomies and practical methods for identifying and comparing risks. Together, these examples show that effective AI governance works best when legal frameworks, international norms, organizational processes, and practical risk tools reinforce one another. The most promising approaches are those that are not only principled, but also operational, adaptable, and usable across sectors and regions.