Obvia - The International Observatory on the Societal Impacts of AI and Digital Technologies

A successful first Global Dialogue on AI Governance would be characterized by concrete, inclusive, and actionable outcomes that move beyond high level principles toward implementation. One key outcome would be the establishment of a shared understanding of core challenges and priorities in AI governance, including safety, human rights, transparency, and societal impacts. This common baseline is essential to facilitate coordination across jurisdictions and sectors while respecting contextual differences. A second outcome would be the strengthening of capacity building and inclusivity. The Dialogue should generate commitments and mechanisms that support the participation of underrepresented regions and stakeholders, particularly from the Global South, as well as civil society, small and medium sized enterprises, and nontechnical experts. This includes access to knowledge, tools, and resources that enable meaningful engagement in governance processes. Third, the Dialogue should produce operational guidance on key challenges. This may include practical frameworks, tools, or reference models to support risk assessment, accountability, human oversight, and the integration of ethical and environmental considerations into AI systems. Such guidance should be adaptable and grounded in real world applications to ensure relevance across sectors. Finally, the success of the Dialogue will depend on its ability to establish a sustained mechanism for continued engagement and follow up. This includes the creation of a structured platform or process that supports ongoing exchange, monitoring of progress, and iterative refinement of governance approaches. Ensuring continuity will be critical to maintaining momentum and enabling adaptive responses to the rapid evolution of AI technologies. Together, these outcomes would position the Global Dialogue as a meaningful step toward more coherent, inclusive, and effective global AI governance.

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

These priorities align closely with the research and action agenda of Obvia - The International Observatory on the Societal Impacts of AI and Digital Technologies, which advances an integrated and responsible approach to AI governance grounded in democratic values. First, the development of "safe, secure and trustworthy AI" (Zicari et al., XXXX) is essential, particularly in high-risk sectors, as it underpins technical reliability, operational safety, and public trust. Second, attention to the social, economic, ethical, cultural, linguistic, and technical implications of AI reflects a systemic perspective, recognizing that AI's impacts extend well beyond the technological domain. In this context, interoperability of governance approaches is essential to ensure coherence across jurisdictions, disciplines, and application settings. Third, the protection and promotion of human rights constitute a central normative pillar, ensuring that AI development and deployment remain aligned with international standards, including non-discrimination, privacy, and equitable access to technological benefits. Finally, transparency, accountability, and human oversight represent key operational conditions for effective governance, ensuring that AI systems remain explainable, auditable, and subject to meaningful human control. Taken together, these four priorities reflect a coherent and complementary framework aimed at reconciling technological performance, social justice, and institutional legitimacy in global AI governance.

The environmental footprint of AI constitutes a major cross cutting issue that requires greater integration into global governance frameworks. In alignment with broader United Nations priorities, particularly those related to climate change, it is essential to systematically account for the environmental impacts of AI systems, including energy consumption, resource use, and associated emissions. This also calls for the promotion of digital sobriety practices and the development of measurement and reporting mechanisms capable of informing more sustainable technological choices. Asymmetries in capacity and infrastructure across countries represent another critical dimension. AI systems operate across borders, yet their development, deployment, and impacts remain unevenly distributed. Variations in regulatory capacity, access to infrastructure, and technical expertise shape how different regions can benefit from or be exposed to the risks of AI. Addressing these disparities requires coordinated efforts to support capacity building, knowledge sharing, and more equitable participation in governance processes. Finally, greater attention should be given to the governance of knowledge and expertise, as well as to the operationalization of human oversight in human-machine interactions. This involves clarifying roles and responsibilities in decision making processes, strengthening interdisciplinary expertise, and ensuring that human intervention remains meaningful and effective throughout the lifecycle of AI systems. Together, these dimensions are essential to advancing coherent, responsible, and globally aligned AI governance.

Governance gaps in the areas of safe and trustworthy AI, human rights, transparency, and societal impacts are already producing tangible effects at both national and international levels. As highlighted by the work of the Observatoire international sur les impacts sociétaux de l'IA et du numérique, the rapid development of AI systems continues to outpace the evolution of regulatory and institutional frameworks, generating persistent uncertainty regarding accountability, risk management, and the protection of fundamental rights. This misalignment weakens public trust and complicates the responsible deployment of AI across sectors. With respect to transparency and human oversight, the absence of harmonized standards limits the capacity of organizations to ensure that AI systems are explainable, auditable, and accountable. This creates concrete challenges for public and private decision makers, particularly in relation to risk assessment, informed decision making, and the implementation of effective accountability mechanisms. These difficulties are compounded by the limited interoperability of governance approaches, which constrains international coordination and reduces policy coherence across jurisdictions. These issues are further amplified by the structural lag between technological innovation and the development of governance frameworks. At the same time, participation remains uneven. Despite recent efforts, the inclusion of civil society, including Indigenous communities, remains insufficiently institutionalized. Strengthening and structuring their participation is essential to ensure that governance processes are more inclusive, representative, and aligned with societal values.

The dialogue plays a critical role in supporting the interoperability of AI governance frameworks by enabling the systematic exchange of best practices, policy instruments, and implementation experiences across jurisdictions. Such dialogue is not limited to states; it involves a broad range of stakeholders, including governments, international organizations, private sector actors, academic institutions, civil society organizations, and technical communities. Through structured and continuous engagement, these actors contribute to aligning regulatory approaches, reducing fragmentation, and fostering mutual learning in a rapidly evolving technological landscape. Beyond coordination, dialogue also serves as a key mechanism for capacity building and inclusivity. It provides a platform through which low- and middle-income countries can access expertise, share context-specific challenges, and participate more meaningfully in shaping global AI governance norms. This is particularly important to mitigate asymmetries in technological development and regulatory capacity, and to ensure that governance frameworks reflect a diversity of perspectives, including those of underrepresented regions and groups. Importantly, AI dialogue must also be understood as a form of scientific diplomacy. By placing scientific knowledge, evidence-based policy, and interdisciplinary collaboration at the center of international exchanges, it helps bridge geopolitical divides and fosters trust among stakeholders. Scientific diplomacy enables the co-construction of shared principles and standards, while supporting the circulation of knowledge across borders and sectors. In this sense, it reinforces the legitimacy and robustness of governance frameworks. Finally, sustained dialogue fosters a more anticipatory and adaptive approach to AI governance. By facilitating early identification of emerging risks and opportunities, and by supporting flexible and iterative coordination mechanisms, it allows stakeholders to respond proactively to technological developments. This forward-looking dimension is essential to ensure that governance remains responsive, inclusive, and aligned with societal values in the long term.

Existing initiatives already provide a strong foundation upon which an AI Dialogue could build. In particular, the work of Obvia illustrates the value of interdisciplinary and multi stakeholder mechanisms by operating as an interuniversity research network that brings together academia, governments, industry and civil society to produce independent and evidence-based knowledge, while supporting democratic deliberation on AI. Its activities include policy-oriented reports, contributions to public consultations, development of ethical tools and participatory initiatives such as webinars, citizen workshops and international collaborations. Several concrete initiatives highlight how these mechanisms can inform an AI Dialogue. Obvia contributes to the UNESCO Readiness Assessment Methodology, which supports countries in aligning national governance frameworks with international ethical standards. It also participates in initiatives such as Z Inspection, recognized by the OECD as a tool for trustworthy AI. In addition, its sectoral work, including the biofood white paper, research on Industry 4.0 and workforce transformation, and studies on AI in higher education, demonstrates how governance can be grounded in real world applications across sectors. Its work on digital literacy, ethical reflexivity and public engagement further strengthens collective capacity and trust. An AI Dialogue would bring added value by connecting and scaling these efforts at the international level. It would support coordination across existing initiatives and contribute to greater coherence between governance frameworks. It would also reinforce scientific diplomacy by placing interdisciplinary research and evidence-based policy at the center of international cooperation, thereby strengthening trust among stakeholders. In addition, it would enhance inclusivity by enabling broader participation from low- and middle-income countries. Finally, it would support a more anticipatory and adaptive approach to governance by facilitating continuous exchange on emerging issues and by encouraging flexible coordination mechanisms.

To be effective, the AI Dialogue should adopt a multi-level and structured format. First, a permanent platform should be established to ensure continuity, supported by thematic working groups reflecting key governance priorities such as ethics, labor, sustainability, and data governance. Second, regular multi-stakeholder forums should enable exchange of best practices and alignment of approaches across jurisdictions. Third, mechanisms for scientific diplomacy should be embedded, ensuring that research, data, and evidence inform discussions and foster trust among stakeholders. Fourth, participatory channels, including citizen consultations and open resources, should be integrated to strengthen legitimacy and inclusivity, building on Obvia's experience with public engagement and knowledge mobilization. Different stakeholders can contribute to the AI Dialogue by mobilizing their complementary expertise, resources, and capacities within a structured and collaborative governance framework. We highlight the importance of interdisciplinary and intersectoral engagement and how partnerships, including in-kind contributions to research initiatives, can concretely support collective action and capacity building. Governments play a central role in convening the Dialogue, ensuring policy coherence, and aligning national frameworks with international standards. Academic institutions contribute by generating independent, evidence-based knowledge and by integrating social, ethical, and interdisciplinary perspectives into AI governance. The private sector brings technical expertise, operational experience, and innovation capacity, while also contributing to the development of standards and responsible practices. Civil society organizations ensure that societal values, human rights, and inclusion remain central, notably through participatory processes and citizen engagement. International organizations can facilitate coordination, provide normative frameworks, and support capacity building, particularly for low- and middle-income countries.

Several voices remain underrepresented in global discussions on AI governance, including Indigenous peoples, youth, civil society and communities directly affected by AI systems, actors from the Global South, non-technical experts, small and medium sized enterprises, as well as women and gender minorities. Addressing these gaps requires moving beyond consultation toward structured, sustained, and context sensitive engagement.

Fostering meaningful and dynamic engagement in an AI Dialogue requires differentiated and innovative formats tailored to the diversity of underrepresented stakeholders. Building on our experience, which combines research, public engagement, and participatory mechanisms, the following approaches can strengthen inclusion and impact. For Indigenous peoples, dedicated co governance circles and culturally grounded dialogue spaces should be established. These formats should recognize Indigenous knowledge systems, support self-determination, and enable sustained engagement rather than one time consultation. For youth, interactive and accessible formats are essential. These may include digital platforms, innovation labs, and youth advisory panels connected to education and AI literacy initiatives, allowing continuous contribution and capacity building. For civil society and communities directly affected by AI systems, citizen assemblies, deliberative forums, and community-based workshops can enable co construction of solutions. These formats should prioritize lived experiences and ensure feedback loops into policy processes. For actors from the Global South, hybrid international forums combined with capacity building hubs can facilitate equitable participation. Twinning mechanisms between institutions and regions can support knowledge exchange and reduce asymmetries in resources and expertise. For non-technical experts, interdisciplinary labs and policy sandboxes can create spaces where social sciences, humanities, and technical perspectives interact. These formats help translate complex issues into actionable governance insights. For small and medium sized enterprises, sector specific roundtables and applied pilot initiatives can support practical engagement. These formats should focus on real use cases, barriers to adoption, and shared solutions. For women and gender minorities, targeted leadership forums, mentorship networks, and inclusive panels can ensure representation and influence. Safe and accessible spaces are essential to support sustained participation.

Several concrete approaches illustrate how AI governance can move from principles to implementation while remaining responsive to societal challenges. Obvia's work provides relevant examples that combine analytical tools, policy support, and engagement mechanisms. First, we recommend ethical reflexivity approaches that enable organizations to integrate social and ethical considerations from the design stage of AI systems. These tools help identify risks such as bias, social impacts, and accountability gaps, while structuring mechanisms for transparency and oversight. They contribute to operationalizing responsible AI within organizations rather than treating ethics as an ex-post consideration. Second, comprehensive analytical reports play a key role in informing governance. Obvia's Status Report on the Societal Impacts of AI synthesizes emerging issues, risks, and opportunities across sectors, offering interdisciplinary insights and concrete recommendations. These reports function as decision support tools for governments by linking scientific knowledge with policy development and by anticipating technological and societal transformations. Third, targeted platforms for policymakers strengthen the translation of knowledge into action. Dedicated forums and colloquia for public decision makers create spaces to discuss regulatory challenges, share experiences, and align approaches. These mechanisms support more coherent and evidence-based policy responses in a rapidly evolving environment. Finally, sustained engagement with civil society is essential to ensure legitimacy and responsiveness. Activities such as public discussions, workshops, and accessible knowledge resources, contribute to integrating diverse perspectives and fostering informed participation in AI governance. These approaches help ground governance frameworks in societal needs and reinforce public trust.