IGF Policy Network on AI

The first Global Dialogue on AI Governance should be judged by whether it can enable meaningful multistakeholder engagement and translate global discussions into coordinated implementation. In particular, it should ensure that stakeholders from developing countries and the Global South are not only represented, but are able to shape priorities, contribute to decision-making, and access the capacity building and financing required to implement AI governance in practice. A core outcome should be stronger alignment across existing global and regional processes. Rather than creating parallel structures, the Dialogue should connect and reinforce ongoing efforts under the AI Summit series, OECD, GPAI, IGF, CSTD, WSIS Forum, ITU AI for Good, and relevant United Nations bodies. This would allow knowledge, policy approaches, and community-driven insights to circulate more effectively across platforms. For example, the International Scientific Panel on AI could present findings to WSIS communities and the global IGF, while inputs from the IGF, PNAI, and NRIs could inform the Dialogue's understanding of local implementation challenges. The Dialogue should also prioritise follow-through by making explicit recommendations on how existing implementation mechanisms can be mobilised. This includes structures under the WSIS and Global Digital Compact processes, such as the UN CSTD Working Group on Data Governance, the International Scientific Panel on AI, and WSIS plus 20 financing mechanisms. Linking outcomes to these institutional pathways would ensure continuity, accountability, and sustained support for countries with limited resources. In doing so, the Dialogue can move beyond high-level discussion and establish a practical, coordinated, and inclusive approach to global AI Governance that is grounded in implementation realities.

• Safe, secure and trustworthy AI
• AI capacity-building
• Interoperability of governance approaches
• Protection and promotion of human rights

Safe, Secure, and Trustworthy AI Ensuring safety and trustworthiness is foundational to public confidence and the long-term sustainability of AI systems. Technical safeguards, resilience, and risk management must be embedded into AI design and deployment. This includes establishing clear "red lines" against harmful practices such as social scoring or subliminal manipulation, while addressing operational risks like misinformation, deepfakes, and data misuse. Trust must be verifiable, relying on auditability, traceability, and independent validation rather than abstract assurances. In this way, AI can serve as a global public good rather than a source of systemic instability. Transparency, Accountability, and Human Oversight Transparency, accountability, and human oversight are the foundation of enforceable AI governance. Without visibility into data, models, and infrastructure, systems risk opacity and unaccountability. Human oversight prevents over-reliance on automation in critical sectors, while verifiable trust through auditability, traceability, and independent validation ensures responsibility. Embedding social, economic, and cultural dimensions within a rights-based framework guarantees AI strengthens human development, aligns with emerging legal standards, and reinforces human rights protections. Interoperability of Governance Approaches Fragmentation across jurisdictions poses one of the most pressing structural challenges. Divergent national and regional frameworks risk creating compliance burdens, regulatory arbitrage, and barriers to innovation. Interoperability does not require uniformity but rather coordination that allows distinct legal systems to function coherently together. Establishing baseline standards and practical mechanisms for cooperation reduces friction, prevents a "race to the bottom," and promotes a globally informed yet locally grounded ecosystem. Such alignment is vital for cross-border collaboration, digital sovereignty, and democratic resilience. AI Capacity-Building Capacity-building is the equity pillar of AI governance, essential to prevent widening inequalities between developed and developing countries. It must go beyond access to compute and datasets, encompassing institutional readiness, governance expertise, procurement tools, and oversight mechanisms. From a rights-based perspective, capacity must align with human rights protections to avoid enabling exclusion or surveillance. Inclusive representation, resource sharing, and institutional strengthening are vital to empower the Global South and marginalized communities. Together, these priorities support a governance approach that moves beyond declaratory principles toward operational oversight, institutional readiness, and sustained multi-stakeholder cooperation.

Several critical dimensions remain insufficiently addressed in current AI governance frameworks. One key area is the intersection with intellectual property (IP) regimes. IP rules shape access to AI technologies, control over datasets and models, and distribution of economic benefits. Ambiguities around authorship, inventorship, and ownership of AI-generated outputs create regulatory uncertainty, while strong proprietary protections risk limiting transparency, accountability, and equitable participation. Governance must balance legitimate rights with public-interest objectives such as auditability, innovation, and capacity-building, particularly in developing countries. Data governance is another strategic priority. Datasets are foundational to AI, yet global asymmetries persist in access and control. Frameworks promoting data-sharing, sovereignty, and public-interest data commons are essential to ensure equitable participation. Closely linked is the recognition of AI as critical digital infrastructure. Compute capacity, cloud services, identity systems, and cybersecurity increasingly determine countries' ability to exercise digital sovereignty. These infrastructure layers must be governed in the public interest, not solely through proprietary control. The political economy of AI also requires attention. Market concentration among a few technology actors creates structural power imbalances, limiting transparency and accountability. Labor impacts are equally pressing, with AI reshaping both formal and informal markets, often displacing workers without adequate protections. Indigenous communities need specific safeguards to prevent appropriation or distortion of cultural knowledge, ensuring respect for civilizational sovereignty and free, prior, informed consent. Gender-based exclusion and exploitation must also be addressed through intersectional governance that mainstreams equality across policy, design, and oversight. Environmental impacts of AI, particularly in the Global South, demand integration into governance frameworks to ensure sustainability. Finally, governance must address non-regression in human cognitive and relational development. Evidence suggests AI-mediated environments can erode empathy, judgment, and social cohesion. Safeguards, monitoring mechanisms, and proportionate compensation models-analogous to environmental governance-should be explored to preserve these capacities as public goods. Together, these cross-cutting issues highlight the need for governance that is equitable, operational, and future-ready, embedding AI within broader sustainable development objectives.

AI governance gaps are affecting regions unevenly, particularly in the Global South, where adoption is advancing faster than oversight systems. This imbalance creates structural vulnerabilities and limits countries' ability to exercise digital sovereignty and align AI deployment with national development priorities. A major challenge is the absence of interoperable governance frameworks. Divergent national regulations complicate cross‑border data flows, compliance, and deployment in sectors such as healthcare, fintech, and public services. Fragmentation reinforces dependency on external systems and hinders alignment with emerging global standards. Limited access to infrastructure—high‑performance computing, quality datasets, and advanced research ecosystems—further restricts domestic innovation, increasing reliance on proprietary platforms and foreign governance conditions. This dependence risks accelerating adoption beyond institutional oversight capacity. Labor market disruption is another critical gap. In many Global South contexts, technological displacement is absorbed through informal labor markets without adequate social protection, masking inequality and leaving independent consultants and contract professionals vulnerable. Governance frameworks must integrate transparency, reskilling, and social protection to address AI‑driven transformations comprehensively. Transparency and accountability deficits are evident in high‑impact use cases such as finance, recruitment, and governance. Without clear auditability and explainability standards, public trust is undermined, exposing individuals to opaque or biased outcomes. Intellectual property and data governance tensions add complexity, as proprietary protections limit transparency and local adaptation, while regulatory ambiguity around AI‑generated outputs creates uncertainty for innovators. Youth face particular risks from AI‑enabled misinformation, unequal access to AI knowledge, and linguistic exclusion. Yet opportunities exist: strengthening AI literacy, embedding youth and gender perspectives, and designing inclusive frameworks can ensure AI supports education, innovation, and social trust. With strategic investments and coordinated global efforts, the AI Dialogue can help bridge governance gaps, enabling responsible innovation while safeguarding rights and equity.

The Global AI Dialogue can serve as a pivotal, neutral, multistakeholder platform under the United Nations, advancing international cooperation on AI governance. Functioning as a network‑of‑networks, it can connect diverse global processes—including the AI Summit series, OECD, GPAI, WSIS Forum, ITU's AI for Good, IGF, PNAI, NRIs, standards bodies, regional institutions, civil society, academia, the technical community, and industry—without duplicating existing forums or centralizing control. The Dialogue can foster trust through transparency, structured knowledge‑sharing, and evidence‑based cooperation. First, it can facilitate convergence around shared principles for safe, secure, and trustworthy AI. While binding harmonization may not be feasible, common standards can reduce fragmentation and support interoperability across governance frameworks. Acting as a translation layer, the Dialogue can help transform principles into practical tools such as audit mechanisms, procurement guidance, and oversight practices adaptable to different legal and development contexts. Second, it can serve as a coordination hub by mapping initiatives, identifying governance gaps, and strengthening linkages with multistakeholder ecosystems such as the IGF and its NRIs, which connect global discussions with local implementation experiences. Third, the Dialogue can advance capacity‑building, particularly for developing countries, addressing inequalities in access to computational infrastructure, datasets, and technical expertise. Inclusive participation should ensure that Global South stakeholders actively shape governance approaches. Fourth, it can convene governments and stakeholders to exchange experiences on cross‑border challenges—including AI safety, cybersecurity, misinformation, data governance, accountability, and socio‑economic impacts—moving discussions toward structured, cooperative, and sustained governance processes. Finally, the Dialogue can serve as a preparatory space for a Global Pact on AI Governance, aligning existing frameworks, human rights obligations, digital sovereignty considerations, and risk mitigation approaches into a coherent global architecture.

The Global AI Dialogue should build upon and connect with existing international initiatives in order to reduce fragmentation and leverage established expertise. Key reference frameworks include UNESCO's Recommendation on the Ethics of AI, the OECD AI Principles, the G20 and G7 processes, including the Hiroshima AI Process, the IGF, the PNAI, the Global Partnership on AI, and the EU AI Act. Regional initiatives, such as ASEAN's AI governance work, also provide important examples of risk-based governance, safety, and accountability frameworks. Particular importance should be given to multistakeholder processes such as the IGF, its NNRIs, and the PNAI, which already convene diverse actors and produce policy-relevant outputs. Their work on governance interoperability, accountability, sustainability, labour impacts, and inclusive digital cooperation offers a strong foundation for the Dialogue. The IGF ecosystem, including the PNAI and the NRIs, could further support the Dialogue as distributed multistakeholder observatory and knowledge-sharing mechanisms, facilitating the exchange of information on governance trends, implementation experiences, and good practices across regions and stakeholder communities. The Dialogue should also connect with technical standard-setting bodies and broader international legal frameworks, ensuring that AI governance is embedded within a wider normative system. Its added value should not be to replicate existing AI ethics principles, but rather to bridge policy, technical, legal, and implementation domains, strengthening coordination and interoperability across initiatives without duplicating them. Importantly, the Dialogue can advance practical convergence through operational tools such as audit methodologies, certification approaches, procurement guidance, AI literacy, cybersecurity safeguards, and capacity-building mechanisms, helping shift AI governance toward verifiable, implementation-oriented, and context-sensitive cooperation, particularly for developing countries and underrepresented regions.

Effective AI governance requires structured and consequential multistakeholder participation under the United Nations that moves beyond symbolic inclusion to ensure accessibility, clear responsibilities, and a direct link between deliberation and implementation. Governance outputs should also be communicated in language intelligible beyond expert communities in order to strengthen legitimacy and public trust. Stakeholder contributions should reflect comparative advantages. Member States provide regulatory perspectives and support interoperable governance approaches. Academia contributes independent analysis and foresight on emerging risks. The private sector offers technical expertise and deployment experience. Civil society highlights public-interest concerns, including human rights and equity. Technical communities and standards bodies develop benchmarks and interoperability standards. Particular attention should also be given to implementation actors—such as public administrations, procurement experts, auditors, and oversight bodies—to ensure that discussions remain grounded in operational realities. The Dialogue should combine inclusiveness with effectiveness through thematic working groups mandated to produce concrete outputs, including policy briefs, model guidelines, audit tools, and readiness benchmarks. Hybrid participation models should combine open consultations, expert roundtables, and regional consultations, particularly involving developing countries and under-resourced institutions. Pairing Member State representatives with non-governmental stakeholders to co-lead sessions could further strengthen multistakeholder balance. High-level governmental segments should remain integrated into multistakeholder plenary discussions, while thematic breakout sessions should systematically feed practical insights back into plenary outcomes. Innovative formats—such as audience-driven discussions, cross-sector matchmaking sessions, and governance "stress-testing" exercises evaluating frameworks in real-world conditions—could further strengthen implementation-oriented cooperation. The Dialogue should operate as a continuous and adaptive mechanism, capable of responding to emerging paradigms such as agentic AI, and supported by transparent reporting and public-facing outputs. Stronger linkages with the IGF, its NRIs, and the PNAI would also be valuable to connect global discussions with regional and national implementation experiences, support inclusive participation, and strengthen shared learning across AI governance processes.

Global discussions on AI Governance continue to reflect structural asymmetries in participation and influence, with several key voices underrepresented despite being directly affected by AI deployment. These include stakeholders from developing countries, local public institutions and regulators, civil society and grassroots actors, indigenous communities, non-English-speaking populations, informal workers, and young people. Many of these actors are responsible for implementation or experience the real-world impacts of AI, yet have limited influence over global governance frameworks. Underrepresentation is not only a question of participation, but also of unequal access to the conditions required for meaningful engagement. Multistakeholder processes such as the IGF demonstrate the value of inclusive dialogue, yet structural barriers persist. Limited technical capacity, financial constraints, language barriers, and restricted access to data, infrastructure, and policy spaces prevent many stakeholders from shaping outcomes, resulting in participation that is often consultative rather than influential. Addressing these gaps requires a shift toward sustained and institutionalised participation across the full policy cycle. This includes targeted capacity-building and financial support, multilingual and regionally grounded engagement, and accessible formats that extend beyond expert communities. Dedicated mechanisms should enable youth, indigenous communities, and affected populations to contribute to agenda-setting, decision-making, and follow-up processes. Inclusion must also recognise diverse forms of knowledge and lived experience, ensuring that governance frameworks reflect cultural and linguistic diversity. Meaningful inclusion should ultimately be measured by the extent to which underrepresented groups are able not only to participate, but to shape outcomes and influence implementation in practice.

Meaningful and dynamic engagement in the AI Dialogue will require moving beyond traditional panel-based formats toward interactive, problem-solving approaches that prioritise co-creation, testing, and real-world applicability. Rather than focusing primarily on exchanging positions, the Dialogue should create structured environments where governance ideas can be developed, challenged, and validated in practice, strengthening the connection between global discussions and implementation on the ground. Formats such as policy labs, co-creation workshops, and multistakeholder "hackathon-style" sessions can enable participants from governments, industry, academia, and civil society to collaboratively develop and stress-test governance frameworks, guidelines, and tools. Scenario-based simulations and AI Governance sandboxes can further support this by allowing stakeholders to work through realistic cases, such as cross-border incidents or AI-driven misinformation, helping to identify gaps in accountability, coordination, and response mechanisms. Participation should not be limited to attendance, but should enable stakeholders to actively shape the Dialogue, contribute to outcomes, and influence decision-making processes. This requires dedicated financial support for participants from developing countries, particularly from the Global South, to ensure equitable and sustained engagement. Inclusive participation can be further strengthened through regional and community-based dialogues, including with the support of National and Regional IGF initiatives, which can bring local perspectives into global discussions and ensure that diverse contexts are reflected in outcomes. Multilingual digital platforms can also expand access and enable broader contributions beyond those physically present. Embedding iterative drafting and feedback cycles across these formats will help ensure that outputs remain transparent, adaptable, and connected across global, regional, and local levels, supporting a more participatory and implementation-oriented governance process.

A range of existing policies and practices offer concrete pathways for effective AI governance. A common pattern across the most effective approaches is the emergence of hybrid models combining regulatory frameworks, voluntary commitments, and practical implementation tools operating across normative, institutional, and technical layers simultaneously. The EU AI Act provides a structured risk-based model, complemented by multistakeholder Codes of Practice, the AI Board, the European AI Office, and the EU AI Pact, which supports early adoption through governance strategies, risk mapping, and AI literacy initiatives. The OECD AI Principles, UNESCO Recommendation on the Ethics of AI, the NIST AI Risk Management Framework, and the Blueprint for an AI Bill of Rights provide complementary approaches to transparency, accountability, fairness, privacy, and human oversight. India's "AI for All" strategy and Digital Public Infrastructure models, including Aadhaar and UPI, demonstrate how interoperable systems can support inclusion when combined with safeguards such as privacy, cybersecurity, transparency, independent oversight, and access to remedy. DPI should therefore be understood not only as scalable technology, but also as public-interest infrastructure. National implementation experiences, including Italy's AgID AI guidelines on adoption, development, and procurement, illustrate how governance principles can be embedded into operational decision-making processes. Emerging practices such as regulatory sandboxes enabling controlled experimentation under supervision, algorithmic audits, impact assessments, open digital ecosystems, data-sharing initiatives, consent architectures, readiness assessments, and structured compliance pathways further help bridge regulatory intent and operational reality. Effective governance should move beyond technical compliance and assess broader societal outcomes, including institutional trust, social cohesion, cultural diversity, democratic participation, and equitable access, supported by continuous monitoring frameworks with indicators and benchmarks. AI literacy, multilingual accessibility, youth engagement, and anticipatory and adaptive governance for emerging paradigms such as agentic AI are also essential for legitimate, environmentally sustainable, and inclusive governance.