Independent expert / Professor of Practice

A successful first Global Dialogue should establish AI governance as a genuinely global, inclusive and implementation-oriented process. It should avoid reducing AI governance either to technical safety alone or to geopolitical competition among technologically powerful actors. AI already shapes education, culture, language, public trust, work, creativity and democratic participation; therefore, its governance must address not only systems, but also the social and cultural conditions in which these systems operate. Three outcomes would be especially important. First, the Dialogue should recognize culture, education, linguistic diversity and human agency as core dimensions of AI governance, not as secondary or "soft" concerns. Smaller languages, local knowledge systems and culturally specific institutions must not become invisible within globally dominant AI infrastructures. Second, the Dialogue should create a shared implementation agenda. Principles are necessary, but not sufficient. The process should encourage practical tools: provenance and consent frameworks, human oversight mechanisms, culturally grounded AI literacy, public-sector guidance, and participatory governance models that can be tested in real settings. Third, the Dialogue should strengthen trust across regions and stakeholder groups by ensuring that developing countries, smaller states, civil society, educators, cultural actors and local communities are not merely consulted, but treated as co-shapers of the agenda. In my view, success would mean that the Dialogue becomes a living bridge between global principles and concrete practice: a space where AI governance is understood as a matter of human dignity, cultural plurality, responsible innovation and the common good.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight
• AI capacity-building

I selected these four priorities because AI governance must address both technical risk and the deeper transformation of human meaning-making, learning and social trust. The social, ethical, cultural and linguistic implications of AI are urgent because AI systems increasingly mediate how societies learn, remember, create, translate, communicate and interpret the world. For smaller languages and culturally specific knowledge systems, the risk is not only exclusion, but also flattening: local meanings may be absorbed into globally dominant statistical patterns without adequate context, provenance or accountability. The protection and promotion of human rights must remain central because AI affects dignity, equality, freedom of expression, access to knowledge, non-discrimination, privacy and cultural rights. Rights-based governance should also include the rights of communities to retain meaningful authorship over their data, memory, heritage and representation. Transparency, accountability and human oversight are essential because AI systems should not become opaque authorities in public life, education, culture, media or public services. Human judgement must remain active, informed and institutionally supported. Oversight should include not only technical audit, but also contextual and cultural review. AI capacity-building is crucial because many societies, institutions and communities are expected to adapt to AI without having sufficient conceptual, technical or governance capacity. Capacity-building should not be limited to technical skills. It should include AI literacy, ethical prompting, institutional readiness, cultural data governance, and the ability to decide when AI should not be used. Taken together, these priorities reflect a simple principle: AI must be governed not only for efficiency and innovation, but for human dignity, plural knowledge systems, cultural diversity and responsible public use.

Yes. One cross-cutting issue deserves stronger attention: narrative and interpretive sovereignty. AI systems do not only process data; they increasingly structure meaning. They influence which stories become visible, which languages are well represented, which forms of knowledge are treated as authoritative, and which cultural contexts are simplified or ignored. For this reason, AI governance should address not only data protection or technical transparency, but also the conditions under which communities, institutions and societies retain agency over their own representation. This is especially important for smaller languages, Indigenous and local knowledge systems, cultural heritage, education and public memory. Cultural materials should not be treated merely as raw material for AI training or content generation. They carry context, dignity, provenance, consent conditions and community meaning. Governance frameworks should therefore include provenance chains, consent and benefit-sharing mechanisms, community review, and the possibility to say no to certain uses. A second emerging issue is the need to move from AI skills to AI wisdom. Many AI literacy frameworks focus on how to use tools effectively. This is necessary but insufficient. Societies also need the capacity to ask why, when, under what limits, and for whose benefit AI should be used. AI literacy should cultivate judgement, reflection, human agency, and awareness of cognitive offloading and automation bias. Finally, environmental responsibility should be treated as inseparable from AI governance. The energy use, hardware demands and e-waste associated with AI must be part of the global conversation. In short, global AI governance should protect not only data and systems, but also meaning, memory, cultural plurality, human judgement and the living conditions of future generations.

In Hungary and the wider Central and Eastern European region, AI governance gaps are especially visible in the relationship between global technological infrastructures and smaller linguistic, cultural and institutional ecosystems. The most significant challenge is not only access to AI, but meaningful participation in shaping it. Hungarian is a relatively small language in global AI development. Even when AI systems function in Hungarian, they may remain weakly attuned to national legal concepts, educational traditions, cultural references, public administration, heritage contexts and local knowledge. This creates risks of linguistic thinning, cultural flattening and interpretive dependency on systems designed elsewhere. A second challenge concerns public institutions. Schools, universities, cultural institutions, museums, archives and public administrations are increasingly expected to use AI, but often lack clear guidance on provenance, consent, human oversight, accountability, procurement and responsible implementation. Without such frameworks, AI adoption may be either overly cautious or too rapid and uncritical. A third challenge is capacity. AI literacy is still too often understood as tool use or prompt technique. The deeper need is for institutional and civic capacity to judge when AI is appropriate, how it should be governed, and how human agency can be preserved. At the same time, the region has important opportunities. Smaller countries can become laboratories for culturally grounded, human-centred AI governance. Hungary has strong traditions in education, culture, science, heritage and linguistic reflection that can support AI systems rooted in context rather than mere scale. Central and Eastern Europe can contribute a distinctive perspective on sovereignty, public trust, institutional resilience and cultural diversity. The key opportunity is to move from passive adoption to responsible adaptation: building AI governance practices that protect human rights, strengthen public institutions, support smaller languages, preserve cultural memory, and turn AI into a companion for learning, creativity and democratic participation rather than a substitute for human judgement.

The AI Dialogue can play a crucial role by becoming a trusted bridge between global principles, regional realities and practical implementation. Many international AI governance discussions remain fragmented: some focus on frontier safety, others on human rights, standards, innovation, development, or sectoral regulation. The Dialogue can help connect these agendas without replacing existing institutions. Its first role should be convening. It can ensure that all countries, including smaller states and developing countries, participate meaningfully in shaping AI governance, not only in adapting to frameworks designed elsewhere. This is essential if AI governance is to avoid reproducing existing digital, linguistic and economic inequalities. Its second role should be translation between communities of expertise. Technical experts, governments, educators, cultural institutions, civil society, the private sector and local communities often use different languages. The Dialogue can create a shared vocabulary for AI governance that connects safety, rights, cultural diversity, capacity-building and public trust. Its third role should be implementation support. The Dialogue should not only produce declarations. It should help identify practical governance tools, model policies, risk assessment methods, capacity-building programmes, and tested examples that countries and institutions can adapt to their own contexts. Finally, the Dialogue can promote mutual learning. AI governance should not flow only from technologically dominant countries outward. It should also learn from smaller languages, local knowledge systems, public institutions, Indigenous communities and culturally grounded practices. Its added value would be to keep AI governance genuinely global, plural and human-centred: a process where technical innovation is guided by human dignity, cultural diversity, democratic accountability and the common good.

The AI Dialogue should build upon existing UN and multilateral initiatives rather than duplicate them. A key foundation is UNESCO's Recommendation on the Ethics of Artificial Intelligence, which provides a globally agreed normative framework grounded in human rights, dignity, inclusion, sustainability and cultural diversity. The Dialogue should connect with UNESCO's practical work on AI readiness, education, media and information literacy, cultural diversity, science and ethics. It should also connect with the Global Digital Compact, the WSIS+20 process, ITU-related digital development and standards discussions, OHCHR's human rights work, WIPO debates on AI and intellectual property, and relevant OECD, Council of Europe, G7, G20 and regional initiatives. These processes contain important expertise, but they are often institutionally fragmented. The Dialogue should also connect with practical educational and cultural initiatives that translate AI governance principles into lived learning environments. One example is the Bartók 4.0 Learning Cycle, a place-based learning methodology from Hungary that links heritage, community, sustainability and responsible AI use through an "analogue first, digital second" sequence. It helps learners move from field encounter and collective interpretation toward ethical digital synthesis, strengthening AI literacy, cultural awareness, human agency and responsible use of AI tools. The Dialogue's added value would be threefold: coherence, inclusion and implementation. It can help different initiatives speak to each other across technical, legal, ethical, cultural and developmental domains; ensure that smaller countries and communities are not only recipients of AI governance, but contributors to it; and identify practical tools, policy templates, repositories of good practice and pilot projects that translate principles into institutional action.

Different stakeholders should contribute according to their distinct responsibilities and forms of knowledge. Governments can provide legitimacy, regulatory experience and public-interest priorities. International organizations can connect existing normative frameworks and capacity-building programmes. The private sector can clarify technical realities, deployment risks and implementation constraints. Academia can contribute independent analysis, evidence and conceptual clarity. Civil society can bring rights-based scrutiny, public accountability and community perspectives. Educators, cultural institutions and local communities can show how AI affects learning, trust, memory, creativity and daily life. The Dialogue should be structured as an ongoing learning process, not only as a high-level event. I would recommend a three-layer format. First, a high-level political and normative segment should define shared priorities and principles. Second, thematic working tracks should address concrete issues such as human oversight, AI literacy, cultural and linguistic diversity, public-sector use, provenance, accountability, capacity-building and environmental sustainability. Third, an implementation and practice layer should collect tested examples, policy tools, risk assessment models, educational methods and pilot projects from different regions. This would allow the Dialogue to learn from practice, not only from declarations. The process should also include regional consultations and stakeholder labs before and after each global meeting. Written submissions should be complemented by short case presentations, structured roundtables and interactive workshops. The Dialogue should avoid becoming either too technical or too diplomatic. Its added value would be to hold these dimensions together: political legitimacy, technical realism, ethical direction and practical implementation.

Several voices remain underrepresented in global AI governance. First, smaller language communities are often absent or treated as secondary. Yet AI systems increasingly shape translation, education, public information and cultural visibility. Languages with fewer digital resources risk becoming functionally present but culturally thinned: available in interface terms, but weakly represented in deeper meaning, context and institutional nuance. Second, cultural and educational actors are underrepresented. Teachers, museums, libraries, archives, heritage professionals, artists and community educators are directly affected by AI, but global AI governance is still often dominated by technical, legal and economic perspectives. These actors understand how AI changes attention, trust, memory, learning and interpretation. Third, Indigenous peoples, local knowledge holders and community-based heritage actors need stronger inclusion. Their knowledge should not be treated as extractable data. Participation must include consent, provenance, benefit-sharing, community review and the right to refuse certain uses. Fourth, young people and women should be included not only as (future) users, but as co-shapers of AI literacy, education and civic imagination. These groups could be included through dedicated stakeholder tracks, regional consultations, multilingual submission channels, community case hearings and funded participation for those without institutional resources. The Dialogue should also support repositories of local and culturally grounded AI governance practices, so that underrepresented communities contribute knowledge, not only testimony. Inclusion should not mean symbolic consultation. It should mean agenda-setting power, feedback loops, and visible incorporation of community concerns into recommendations, tools and future workstreams. The basic principle should be clear: those whose languages, cultures, data, labour, creativity and learning environments are shaped by AI must have a meaningful role in shaping AI governance.

The AI Dialogue should combine formal deliberation with more dynamic formats that allow participants to learn from real cases, tensions and implementation challenges. One useful format would be "AI governance clinics," where countries, institutions or communities present a concrete governance dilemma and receive structured feedback from technical, legal, ethical, educational and cultural experts. This would move the Dialogue from abstract principles to practical problem-solving. A second format could be "regional practice labs," organized before and after the global meetings. These labs could collect examples of AI governance in education, culture, public administration, health, media, open-source development and local innovation. Their results could feed into the global Dialogue as living evidence. A third format would be "scenario roundtables." Participants would examine plausible future cases: AI in schools, AI-generated public disinformation, automated translation of minority languages, AI use in cultural heritage, or public-sector decision support. This would help identify governance gaps before harms become irreversible. A fourth format could be a curated "implementation repository" linked to the Dialogue. It would gather policy templates, consent models, procurement clauses, AI literacy curricula, audit tools, provenance frameworks and tested pilot projects. This would make the Dialogue useful beyond the meeting itself. The Dialogue could also include short "community testimony sessions" and youth-led panels, especially for smaller languages, Indigenous knowledge, education, culture and civil society. Finally, it should encourage demonstrations of responsible AI practices, not only presentations about AI. Examples such as place-based learning, analogue-first AI literacy, community-reviewed digital companions, and culturally grounded governance pilots could show how principles become practice. The most effective format would therefore be hybrid: diplomatic enough to build consensus, practical enough to support implementation, and participatory enough to learn from those most affected by AI.

Several existing approaches can help translate AI governance principles into practice. First, UNESCO's Recommendation on the Ethics of Artificial Intelligence offers a strong normative foundation, especially through its focus on human rights, dignity, inclusion, sustainability, cultural diversity and policy readiness. Its value lies in connecting principles with implementation tools for Member States and institutions. Second, AI governance should build on human rights impact assessments, algorithmic impact assessments, risk registers, audit mechanisms and responsible public-sector procurement clauses. These tools can help ensure that AI systems are transparent, accountable, proportionate and subject to meaningful human oversight before deployment. Third, AI literacy initiatives should move beyond technical skills. Effective practice should cultivate judgement, critical thinking, awareness of automation bias, cognitive offloading, data provenance and the ethical limits of AI use. The Bartók 4.0 Learning Cycle offers a useful example: learners begin with field encounter, observation and community interpretation before using AI for ethical digital synthesis. This "analogue first, digital second" sequence helps preserve human agency and contextual understanding. Fourth, the "Ethical AI Companions for Cultural Heritage" case from Hungary, featured by the UNESCO International Centre for Creativity and Sustainable Development, provides a practical example of culturally grounded AI governance. It connects Pannonhalma, Lake Turkana (Kenya) and Tihany (all World Heritage sites) through a model in which AI is used as a supervised learning and interpretation companion, not an autonomous decision-maker. The approach emphasizes community authorization, human review, cultural rights, narrative sovereignty, and UNESCO's AI ethics principles.