UNESCO

A successful first Global Dialogue on AI Governance should be judged by its ability to move from broad exchange to shared direction for action. It should generate clear, politically supported priorities for international cooperation, including capacity-building, institutional safeguards, interoperable governance approaches, supervisory cooperation, and support for countries with limited resources. A key success factor would be for the Dialogue to be informed and anchored by existing global normative instruments, in particular UNESCO's Recommendation on the Ethics of Artificial Intelligence. As the first global normative instrument on AI ethics, adopted by all UNESCO Member States, the Recommendation can help level the playing field, define the remit of the Dialogue, and align discussions by design around human rights, dignity, inclusion, gender equality, cultural and linguistic diversity, environmental sustainability and the public good. Success would also require a strong interface with the Independent Scientific Panel on AI, so that discussions are informed by credible, multidisciplinary evidence and horizon scanning. This should help participants distinguish real risks and opportunities from hype and identify priorities for collective action. The Dialogue should also connect global discussion with implementation realities. UNESCO's trusted relationships with AI-related institutions, public authorities, expert networks, civil society, academia, media, cultural actors and other stakeholders across regions can help keep the Dialogue grounded in what is feasible in practice. This is essential for moving from discussion to action in an efficient, inclusive and credible way. Finally, the Dialogue should establish practical follow-up mechanisms and workstreams, including on media and information literacy, linguistic diversity, multilingual technologies, and capacity-building, so that the Geneva and New York discussions lead to sustained cooperation and tangible benefits for people.

• Safe, secure and trustworthy AI
• AI capacity-building
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights

UNESCO's priorities for urgent action and active engagement under the Global Dialogue on AI Governance align most strongly with the thematic areas of AI capacity-building; protection and promotion of human rights; transparency, accountability and human oversight; and interoperability of governance approaches. AI capacity-building is a foundational priority. Without adequate skills, institutions, especially small and medium-sized media outlets and under-resourced public bodies, risk exclusion, loss of viability and weakened pluralism. Strengthening digital and AI competencies across media institutions, civil servants, educators and judicial actors is essential to ensure that AI supports, rather than undermines, independent journalism, inclusive public services and access to justice. UNESCO addresses these needs through global capacity-building agendas, including its Digital Competency Framework for Civil Servants and AI Competency Frameworks for teachers, students and policymakers. Through the Internet for Trust Initiative and the Guidelines for the Governance of Digital Platforms, including their generative AI companion, UNESCO promotes a human rights-based, risk-based and multistakeholder approach to platform and AI governance. Priority concerns include freedom of expression, fair compensation for journalistic and artistic content used in AI training, risks of disinformation and hate speech, underscoring the importance of information integrity, and cultural and linguistic bias in algorithmic systems. Transparency, accountability and human oversight are particularly urgent in high-stakes public-sector uses of AI, notably in the justice system, where opaque systems may undermine due process, equality before the law and public trust. Interoperability of governance approaches is equally critical. AI governance increasingly operates across borders and institutional systems. UNESCO fosters peer exchange, communities of practice and shared methodologies, such as through global regulator forums and knowledge networks, to reduce fragmentation while enabling locally adapted solutions anchored in common values. These priorities directly support the implementation of UNESCO's Recommendation on the Ethics of Artificial Intelligence, ensuring that "safe, secure and trustworthy AI" is understood as ethical, rights-based, inclusive and socially grounded rather than narrowly technical. Finally, open-source software, open data, and open AI models are essential prerequisites for safe, secure, and trustworthy AI, as they enable transparency, independent scrutiny, reproducibility, and broad scientific collaboration while reducing concentration of power over critical technologies. These principles are consistent with the UNESCO Recommendation on Open Science (2021), which emphasizes openness, accessibility, inclusiveness, and international cooperation as foundations for trustworthy and equitable scientific and technological development.

Institutional readiness and organisational transformation in public administration and justice systems remain under-addressed. Beyond skills gaps, many governments and judiciaries face structural challenges in embedding AI responsibly, including misaligned governance arrangements, leadership, incentives and funding models. UNESCO's work shows that AI initiatives often fail when institutional mandates, coordination mechanisms and change-management processes do not match the complexity of AI technologies. Addressing AI governance therefore requires sustained attention to public-sector reform, civil service training ecosystems, cross-ministerial coordination and long-term capacity-building. Data governance should be recognised as a foundational pillar of AI governance. AI systems are built on data, and gaps in how data are collected, accessed, shared and governed directly lead to bias, exclusion, loss of trust and power asymmetries. Treating data governance as a downstream or technical issue limits the effectiveness of algorithmic regulation. Rights-based data governance is essential for inclusive and accountable AI, and for safeguarding the independence, integrity and effectiveness of rule-of-law institutions. A related emerging concern is the impact of AI on judicial independence and constitutional guarantees, with risks of creating dependency on proprietary systems, amplify historical bias, or subtly shift authority from judges to technical systems and vendors. These risks need focused attention beyond general transparency or human-rights discussions. A cross-cutting theme could include "Institutional safeguards for AI in public administration and justice systems." The protection of cultural rights deserves stronger visibility. AI systems are trained on vast cultural and heritage datasets, often without consent, raising issues of cultural sovereignty, attribution and equitable benefit-sharing. At the same time, AI offers opportunities for heritage preservation and language revitalisation that require rights-based governance to ensure community control. Finally, there is a persistent implementation gap between normative convergence and institutional capability. Values such as human rights, transparency and inclusion are widely shared, but unevenly operationalised. Treating "implementation readiness" as a cross-cutting issue would help bridge this gap and translate global commitments into institutional practice.

AI governance gaps are having significant and uneven impacts across regions and sectors, particularly where institutional capacity, regulatory frameworks and digital infrastructure lag behind rapid technological change. For Indigenous Peoples, AI presents both important opportunities and serious risks. While AI can support language revitalisation, knowledge transmission and cultural visibility, weak data governance and unclear rules on consent, ownership and benefit sharing expose Indigenous knowledge systems to misappropriation and data exploitation, undermining cultural sovereignty. UNESCO therefore prioritises advancing Indigenous rights in digital governance, including data governance, intellectual sovereignty and protection of traditional knowledge. In the media sector, the gap between fast moving AI innovation and slow regulatory adaptation is exacerbating structural inequalities. Large media organisations are better positioned to adopt AI, while smaller outlets face legal uncertainty, high costs and a widening verification gap as synthetic content outpaces editorial capacity. These trends also heighten information integrity risks, reinforcing the need for content provenance tools, stronger verification capacities and support for trusted public-interest media. The absence of content provenance standards and algorithmic transparency shifts power toward platform infrastructures, incentivising engagement over quality, increasing ethical risks and threatening media pluralism. Across public administration, fragmented governance frameworks and limited institutional capacity often lead to ad hoc AI deployments with insufficient attention to human rights, accountability and sustainability. This can deepen digital divides, reinforce unequal access to public services and erode public trust. Resource intensive AI models further compound inequalities, especially in low resource contexts. UNESCO's work on Green AI demonstrates that resource efficient, right sized AI is essential for inclusive and feasible governance. At the same time, capacity building presents a major opportunity. Governments that invest in AI literacy and competency development for civil servants are better able to steer AI projects toward inclusive service delivery, data protection and meaningful human oversight. In the justice sector, AI tools offer efficiencies in legal research and case management, but weak governance can threaten due process, equality before the law and fair trial rights. Conversely, targeted training and guidance have improved judicial awareness and safer, context sensitive innovation. Finally, AI governance gaps are acutely felt in the culture and creative industries. Legal uncertainty around training data, copyright and AI generated content undermines livelihoods, while algorithmic amplification marginalises local, minority and Indigenous cultural expressions, challenging cultural diversity objectives.

The AI Dialogue can play a decisive role in advancing international cooperation by serving as a practical bridge between global norms and the everyday governance of AI in public administration, justice and other core public-interest sectors. Its value lies in aligning shared standards with concrete implementation, capacity-building and peer cooperation across countries and regions. In this regard, UNESCO's Recommendation on the Ethics of Artificial Intelligence can serve as a common normative anchor for the Dialogue, helping to align cooperation around shared values while leaving room for diverse national and regional governance approaches. First, the Dialogue can provide a global space to align existing international norms, including human rights, rule of law and ethical AI standards, with operational governance practices. By focusing on how countries translate principles into legislation, institutional safeguards and oversight mechanisms, it can help move cooperation from consensus on values to convergence in practice. Second, the Dialogue can coordinate and amplify AI capacity building for public institutions by promoting shared tools, curricula and competency frameworks, particularly for civil servants, regulators and judicial actors. This is essential to reduce fragmentation, avoid duplication and ensure that countries with limited resources can meaningfully participate in AI governance. Third, the Dialogue can surface and connect good practices on rights based AI use in public services and justice systems, including procurement safeguards, impact assessment, human oversight and institutional reform. Peer exchange across legal systems and development contexts can support locally adapted solutions anchored in common principles. Fourth, it can foster interoperability of governance approaches by encouraging shared methodologies and learning processes that respect different legal traditions while enabling cross border cooperation. Finally, the Dialogue can function as a clearinghouse between scientific evidence and implementation experience, linking insights from the Independent Scientific Panel with lessons emerging from existing tools and country-level practice, including readiness assessments, ethical impact assessments and capacity-building initiatives. In doing so, it can help shape shared policy questions, targeted cooperation priorities and sustained international learning, while complementing, not duplicating, existing normative frameworks.

The AI Dialogue should build on and connect existing normative, programmatic and multistakeholder initiatives that already advance inclusive, ethical and rights-based AI governance. Key UNESCO foundations include the Recommendation on the Ethics of Artificial Intelligence (2021), the Recommendation on Open Science (2021), and the Recommendation on Multilingualism and Universal Access to Cyberspace (2003); the International Decade of Indigenous Languages (2022–2032), "The guidelines for the governance of digital platforms and generative artificial intelligence: companion document", and related cultural policy frameworks such as MONDIACULT outcomes and work on AI and cultural diversity. the 2021 UNESCO Recommendation on Open Science, Operational initiatives such as UNESCO's Global Roadmap for Language Technologies, World Atlas of Languages, and the WSIS Action Lines, notably Action Line 8 on cultural and linguistic diversity, are also highly relevant. Among these, UNESCO's Recommendation on the Ethics of Artificial Intelligence should be recognised not only as a reference document, but as a global normative instrument capable of informing the Dialogue's scope, framing and follow-up workstreams. The Dialogue should connect with wider global mechanisms, including the Global Digital Compact, ITU's AI for Good, OECD AI policy work, and open-source and digital public goods ecosystems that support multilingual data, models and infrastructure. Within UNESCO, valuable implementation ecosystems already exist around digital platform governance and AI, including the Global Forum of Networks of Regulators, the Internet for Trust Knowledge Network, and multistakeholder convenings such as the International Conference on Digital Platform Governance. In the public sector and justice systems, the Dialogue can build on UNESCO's Programme on AI for the Public Sector, the Digital Competency Framework for Civil Servants, the SPARK‑AI Alliance, and the AI and the Rule of Law Programme, including global toolkits and MOOCs for judges and legal professionals. It should also connect with the implementation ecosystem of the AI Ethics Recommendation, including the Readiness Assessment Methodology (RAM), Ethical Impact Assessments (EIA), the Global AI Ethics and Governance Observatory, and associated expert, civil society and private‑sector networks. The added value of the AI Dialogue lies in connecting these existing assets, avoiding duplication, and linking norms, scientific evidence and implementation experience. By acting as a coordination hub across UN processes and communities of practice, the Dialogue can strengthen interoperability, promote multilingual and open access approaches, and translate global principles into sustained international cooperation and practical governance capacity.

For the AI Dialogue to produce practical and credible outcomes, public administrations and judiciaries should be recognised as core stakeholders. These actors are directly responsible for designing, procuring and deploying AI in essential state functions and for safeguarding human rights, due process and the rule of law. Their operational experience with real users, legacy systems and institutional constraints is essential for shaping governance approaches that are workable in practice, not merely aspirational. To enable meaningful contributions, the AI Dialogue should move beyond statement‑driven plenaries and adopt interactive, problem‑solving formats, such as policy labs and case clinics. In these sessions, mixed groups, public officials, judges, regulators, civil society, culture stakeholders and technical experts, would work through concrete use cases (e.g. AI in eligibility assessments, court case management, legal research, or generative AI in cultural production), jointly identifying risks, safeguards and institutional arrangements. Systematically documenting the outcomes would help distil adaptable governance practices and templates, aligned with instruments such as the UNESCO Recommendation on the Ethics of Artificial Intelligence, while building sustained communities of practice across regions. Culture stakeholders, including artists, cultural institutions, heritage bodies, creative industry organisations and civil society, should also be explicitly integrated into the Dialogue's structure. AI significantly affects cultural diversity, creative livelihoods and heritage preservation, yet these perspectives are often under‑represented in AI governance discussions. Dedicated thematic sessions on AI and culture, moderated roundtables and reporting mechanisms should ensure these contributions meaningfully inform outcomes. Finally, the Dialogue could be organised around existing governance tools. RAM‑based policy labs could examine institutional readiness; EIA‑based case clinics could test safeguards for concrete AI deployments; and supervisory exchanges could compare oversight models. This approach would make the Dialogue less performative and more operational, focusing on how AI governance actually functions in practice.

Several voices and perspectives remain significantly underrepresented in global discussions on AI governance, despite being among those most affected by AI systems and decisions. Indigenous Peoples are foremost among these groups. AI systems increasingly draw on data that may include Indigenous languages, cultural expressions and traditional knowledge, often without consent or appropriate safeguards. Exclusion from governance discussions risks perpetuating historical patterns of marginalisation and undermining Indigenous data sovereignty. Inclusion requires the direct participation of Indigenous representatives in policy forums, support for Indigenous‑led AI initiatives, and governance frameworks that protect cultural heritage, languages and traditional knowledge while enabling AI to support revitalisation rather than erosion. Minority and low‑resource language communities, as well as persons with disabilities, are also underrepresented. Many AI systems fail to support accessible and multilingual communication tools such as speech‑to‑text, captioning or sign‑language technologies, reflecting their absence from training data, benchmarks and design processes. Communities facing limited connectivity and digital infrastructure are further excluded from benefiting from or shaping AI systems. More broadly, actors from the Global South, grassroots civil society, independent researchers, local and subnational authorities, frontline public servants, judges and judicial training institutions often remain "invited but unheard." While they may participate in consultations, they frequently lack the influence, resources or institutional pathways to shape outcomes, despite facing disproportionate risks from AI deployment in public services, justice systems and cultural sectors. Culture professionals and practitioners, including artists, heritage institutions and traditional knowledge holders, are similarly underrepresented, even as AI transforms cultural production, distribution and monetisation and raises risks of appropriation and loss of cultural diversity. To address these gaps, the AI Dialogue should proactively reserve space and resources for underrepresented groups, particularly from low‑resource contexts; support early and sustained participation across diagnostics, policy design and impact assessment; and establish transparent feedback loops showing how inputs shape outcomes. Inclusion should move beyond representation toward meaningful co‑design of AI governance solutions grounded in lived experience and institutional practice.

Innovative engagement formats for the AI Dialogue should move beyond traditional plenary‑driven consultations to foster inclusive, practical and dynamic participation. Formats should prioritise co‑creation, problem‑solving and multilingual access, ensuring that engagement leads to concrete governance insights rather than symbolic participation. A first priority is multilingual and culturally responsive participation. The Dialogue could include multilingual deliberative labs, hybrid regional listening sessions with real‑time interpretation, including for low‑resource languages, and community‑led roundtables facilitated by Indigenous Peoples and minority language users themselves. Participatory digital platforms enabling asynchronous input in multiple formats (written, audio, visual and oral) would further reduce barriers linked to language, literacy and connectivity, and support broader global participation. Second, the Dialogue should adopt practice‑oriented formats such as "AI in practice" labs or case clinics. In these smaller, mixed‑stakeholder groups, participants could work through concrete AI use cases, spanning public services, justice, media or culture, jointly identifying risks, safeguards and institutional governance options grounded in human rights and rule‑of‑law principles. Smaller group settings also help ensure that voices less inclined to speak in plenary sessions are meaningfully heard. Third, scenario‑based and simulation exercises could help participants explore complex trade‑offs in AI governance. Policymakers, regulators, industry, civil society, technical experts and affected communities could work through shared scenarios from different perspectives, then reconvene to compare approaches and identify areas for convergence and cooperation. Finally, the Dialogue would benefit from the involvement of professional facilitators and design‑thinking experts. Their role would be to structure interactions, ensure balanced participation, and translate discussions into documented outcomes. This would enhance the legitimacy, inclusiveness and effectiveness of the Dialogue, positioning it as a space where stakeholders actively shape workable AI governance solutions.

Several existing policies, practices and platforms offer concrete, transferable solutions for effective and inclusive AI governance. One example is UNESCO's Dive into Heritage (DIH) project, which uses advanced digital technologies to document and present World Heritage sites through an immersive online platform. DIH combines international standards for digital data collection, strong protection of digital intellectual property through tailored data-collaboration agreements, and a sustained capacity-building programme that has trained more than 100 site managers and professionals. This integrated approach demonstrates how AI-enabled technologies can support cultural preservation while respecting data ownership and rights. Another example is the Intellectual Property and Culture Subgroup of UNESCO's Civil Society Organizations and Academic Network on AI Ethics. This multistakeholder platform brings together civil society, academia and experts to address AI's impact on creative ecosystems, cultural production and knowledge commons. Its work, including a policy brief on gaps in intellectual property regimes in the AI context and the development of a resource hub for creators, offers rights-based, gender-responsive guidance to protect cultural sovereignty, freedom of expression and artists' livelihoods, while addressing risks such as the non-consensual use of cultural and Indigenous knowledge. UNESCO's long-standing work on multilingualism, Indigenous languages and data governance also provides a strong governance model. Instruments such as the 2003 Recommendation on Multilingualism, the 2021 Recommendation on the Ethics of Artificial Intelligence, and the International Decade of Indigenous Languages promote inclusive, rights-based approaches that embed Indigenous data sovereignty and full community participation in the design and deployment of AI systems. Most comprehensively, UNESCO's implementation ecosystem for the Recommendation on the Ethics of AI offers a practical governance framework that links norms to action. By combining country diagnostics (RAM), system-level assessment (EIA), open knowledge platforms, expert support, sectoral toolkits and structured stakeholder networks, this approach treats AI governance as a continuous implementation system. It provides a concrete model for international cooperation that the AI Dialogue could build upon to turn shared principles into operational practice across diverse contexts. It provides a concrete model for international cooperation that the AI Dialogue could build upon: a trusted implementation pathway linking global norms, national institutions, expert networks, stakeholders and practical tools, so that shared principles can be translated into operational practice across diverse contexts.