German University of Digital Science

From our perspective as the founders of the German University of Digital Science — Europe's first fully digital university — the inaugural Global Dialogue on AI Governance (Geneva, July 2026) would be a genuine success if it delivers five concrete outcomes: A deliverable-oriented roadmap, not a declaration. Success means a structured, multi-year action plan with named deliverables, responsible actors, and review cycles — not another statement of principles. Governance is already forming through practice, in fragmented national and regional tracks. The Dialogue must provide the coordinating spine before divergence becomes irreversible. Sector-specific guidance on AI in education. A dedicated workstream on educational AI — with initial standards on learner audit rights, transparency in AI-assisted assessment, and minimum human-oversight requirements for pedagogical decisions — would signal that the Dialogue treats high-stakes AI domains with the specificity they demand. Structural inclusion of Global Majority and academic voices. The 89-country New Delhi Declaration demonstrated that the Global South has substantive, coordinated positions. Universities and civil society must be recognised as contributors, not consultees. Regulatory interoperability. Baseline alignment between the EU AI Act, national frameworks, and the emerging UN architecture reduces compliance fragmentation for institutions and developers operating across jurisdictions. Recognition of educational sovereignty. AI is increasingly an epistemic actor — shaping what is learned, assessed, and valued. The Dialogue should formally acknowledge this as a distinct governance challenge, irreducible to data protection or market regulation.

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

As founders of Germany's first fully digital university, the German University of Digital Science, our selection of these four priorities reflects both our institutional identity and our frontline experience with AI in higher education. (i) Safe, Secure and Trustworthy AI is our first priority because AI systems in education - adaptive tutors, automated assessment, student analytics - directly affect learners' cognitive development, data privacy, and fundamental rights. A fully digital university cannot remain neutral on safety standards: every instructional design decision involving AI carries trust implications that existing governance frameworks do not yet adequately address. (ii) AI Capacity-Building matters to us across three levels simultaneously: the global North-South divide in AI infrastructure, the intra-national gap between well-resourced and under-resourced institutions, and the urgent need for teacher and student AI literacy at scale. Our MOOC-based, internationally oriented model positions us as an active contributor to closing these gaps - not merely a beneficiary of global capacity programmes. (iii) Social, Economic, Ethical, Cultural, Linguistic and Technical Implications aligns directly with our Digital Science mandate. The linguistic dimension is particularly acute: AI systems predominantly trained on English-language data structurally disadvantage non-English learners - a systemic inequity that governance discussions rarely name explicitly. (iv) Transparency, Accountability and Human Oversight translates a pedagogical principle into a governance demand: teachers and institutions - not algorithms - must remain accountable for decisions shaping learners' lives. We call for enforceable disclosure obligations, learner audit rights, and minimum human-oversight standards for all AI used in pedagogical decision-making.

AI in Education: An Urgent Priority for the Global Dialogue on AI Governance The Global Dialogue on AI Governance, established by A/RES/79/325, rightly centres three priority clusters: safe and trustworthy AI, equity and capacity-building across the digital divide, and openness and interoperability of governance frameworks. AI in education sits at the intersection of all three - and demands urgent dedicated attention. As founders of the German University of Digital Science, Germany's first fully digital university, we witness firsthand how AI is restructuring the conditions of learning, teaching, and institutional decision-making at unprecedented speed. Three imperatives emerge from this vantage point: First, safety and accountability in learning environments. AI-driven adaptive learning, automated assessment, and student data analytics must be subject to meaningful human oversight and transparency standards - the resolution's own call for "human-centred" governance must extend explicitly into educational institutions. Second, closing the educational dimension of the digital divide. Capacity-building provisions in the resolution focus on developing countries - but within all societies, including Germany, AI literacy gaps between institutions, social groups, and generations reproduce structural inequalities. Universal AI literacy must be treated as a public good, anchored in higher education policy. Third - the cross-cutting gap: The current thematic framework does not explicitly address AI as a pedagogical and epistemic actor. When AI shapes what students learn, how they reason, and what counts as knowledge, governance is no longer merely about data protection or market access. It becomes a question of educational sovereignty - who controls the epistemic infrastructure of future generations. This emerging issue deserves a dedicated thematic strand in the 2026 Geneva Dialogue.

The governance gaps in A/RES/79/325's thematic architecture directly translate into four compounding challenges — and corresponding opportunities — for higher education worldwide. The safety gap creates institutional liability without protection. AI tools for assessment, admissions, and adaptive learning are proliferating faster than any safety standards exist to evaluate them. Universities deploy these systems in legal and normative vacuums, bearing full accountability while having no internationally recognised benchmarks to validate their choices. The opportunity: sector-specific safety standards for educational AI — modelled on the proposed Global Frontier AI Evaluation Framework — would give institutions actionable compliance criteria and rebuild learner trust. The capacity gap reproduces structural inequality inside education. AI literacy is becoming a baseline professional and civic competency, yet access to AI-capable faculty, infrastructure, and curricula remains deeply unequal — between Global North and South, and within advanced economies between elite and under-resourced institutions. UNESCO's AI Competency Frameworks for teachers and students are important instruments, but lack binding implementation mechanisms. The opportunity: the Dialogue can mandate capacity-building programmes that treat universities — not just governments — as primary delivery partners. The transparency gap undermines academic integrity. Without enforceable disclosure obligations, AI-generated content, AI-assisted grading, and algorithmic admissions decisions erode the epistemic foundations of academic life. Learners cannot contest what they cannot see. The opportunity: the Dialogue could establish a first global standard for AI transparency in educational assessment — a breakthrough for academic integrity globally. The thematic gap on epistemic implications is the deepest structural risk. None of the Dialogue's current themes explicitly address AI as a knowledge authority — a system that increasingly determines what students encounter, how they reason, and what counts as credible evidence. For higher education, this is not peripheral: it strikes at the core of institutional autonomy, curriculum sovereignty, and academic freedom. The opportunity — and the urgency — is for the Geneva session to open precisely this conversation

The Global Dialogue on AI Governance is structurally unique: the first universal, multi-stakeholder forum in which all 193 UN Member States, academia, civil society, and industry engage on AI governance as peers. As founders of Europe's first digital university, we see three roles as decisive. Anchoring AI education within global governance. AI education — teacher and student AI literacy, curriculum sovereignty, equitable access to AI-capable institutions — is currently absent as an explicit agenda item in any global governance forum. The Dialogue can change this by establishing a dedicated workstream on AI in education, producing the first internationally recognised standards for safe, transparent, and human-overseen AI in learning environments. The 30 April 2026 written-input deadline is the immediate entry point. Bridging fragmented governance regimes. AI governance is forming in parallel silos — the EU AI Act, the OECD Principles, the 89-country New Delhi Declaration. No existing forum has the universality to align these tracks. The Dialogue can establish a Regulatory Interoperability Mechanism preventing any single jurisdiction from setting the global standard by default — essential for universities operating across borders. Connecting the Independent Scientific Panel to educational policy. The parallel AI Scientific Panel — 40 multidisciplinary experts — can assess AI's impacts on learning outcomes, academic integrity, and epistemic development across regions. The Dialogue is the mechanism by which those findings become governance commitments, creating the science-policy interface that educational AI currently lacks entirely. Recognising universities as substantive contributors — not consultees — is the condition for this to work

The AI Dialogue should not start from scratch — a rich ecosystem of existing initiatives already provides the building blocks. Its added value lies in converting these fragmented efforts into binding, universal commitments. Existing initiatives to build upon: UNESCO's AI Competency Frameworks for teachers and students (2024) define 12 competencies across four dimensions — human-centred mindset, AI ethics, techniques, and system design — across three progression levels. They are the most comprehensive educational AI literacy standard available, but remain non-binding guidance without institutional enforcement mechanisms. OECD/EC AILit Framework and OECD Digital Education Outlook provide governance guidance for generative AI in education across 18 countries and feed directly into PISA 2029's Media and AI Literacy assessment. The OECD AI Principles — the first intergovernmental AI standard, endorsed by the G20 and 46+ countries — have already shaped the EU AI Act and national legislation worldwide. ITU AI Skills Coalition (launched Davos 2025), with 25+ founding partners including AWS and Microsoft, aims to democratise AI training globally — directly addressing the capacity gap. UNICEF-ITU Giga Initiative works to connect every school globally by 2030, using AI to map infrastructure gaps.

The Dialogue's founding resolution explicitly designs it as a multi-stakeholder platform — governments, academia, civil society, industry, and technical communities all have defined roles. From the perspective of the German University of Digital Science, meaningful contribution requires both the right structural design and differentiated stakeholder mandates. Stakeholder contributions: Governments set the governance agenda and translate Dialogue outcomes into national legislation and bilateral commitments. Their core contribution is political will — endorsing interoperability mechanisms and capacity-building funding. Universities and research institutions provide the independent evidence base that neither industry nor governments can credibly supply. They should feed directly into the Independent Scientific Panel's work, contribute education-specific impact assessments, and co-design sector guidance. Civil society and learner organisations represent those most affected by educational AI — students, teachers, marginalised communities — and must have formal speaking and review rights, not merely observer status. Industry contributes technical knowledge and implementation capacity, but its participation must be structured to prevent regulatory capture — through conflict-of-interest disclosure requirements and separation between deliberative and advisory roles. Format and structure recommendations: The Dialogue should operate in three interlocking tiers: plenary sessions for political endorsement of outcomes; thematic working groups — including a dedicated AI-in-education strand — for substantive standard-setting; and intersessional consultations (like the 11 March 2026 virtual session) enabling year-round written input and civil society engagement. Annual alternation between Geneva and New York preserves universality while connecting to both the ITU AI for Good infrastructure and the UN General Assembly cycle. Outcomes should be graduated — distinguishing non-binding guidelines from endorsed standards from mandated reviews — to maintain political consensus while creating meaningful accountability.

Five communities remain structurally underrepresented in global AI governance discussions, each requiring specific inclusion mechanisms. Learners and students. Those most directly affected by educational AI — the epistemic choices it makes on their behalf — have no formal representation in any governance forum. A dedicated youth and learner advisory panel, with formal speaking rights in the education workstream, would address this directly. Global South researchers and institutions. The New Delhi Declaration demonstrated coherent Global South positioning, yet AI governance standard-setting remains concentrated in Washington, Brussels, and Geneva. Rotating co-chairs from underrepresented regions, funded participation, and regional preparatory consultations are minimum requirements. Non-English linguistic communities. AI systems are predominantly trained on English-language data, and governance deliberations are conducted almost exclusively in English. Multilingual documentation, interpretation, and — critically — multilingual AI assessment tools must be treated as governance infrastructure, not optional services. Indigenous and traditional knowledge communities. Their knowledge systems are both threatened by AI's homogenising tendencies and absent from AI training datasets — a dual invisibility with no current remedy in any governance framework. Teachers and frontline educators. UNESCO's competency frameworks address teachers as objects of upskilling, not as governance actors. Structured teacher representation — through national education unions and professional associations — in the Dialogue's education workstream would correct this imbalance.

Drawing on the German University of Digital Science's experience as a fully digital institution, the following formats would move the Dialogue beyond traditional intergovernmental procedure toward genuinely dynamic, inclusive deliberation. Living Labs and real-time demonstrations. Rather than discussing AI governance abstractly, sessions should embed live demonstrations of AI tools — in education, healthcare, public administration — where delegates experience the governance challenges firsthand. The ITU AI for Good Summit model already integrates technical showcases with policy deliberation; the Dialogue should systematise this into every thematic workstream. Structured stakeholder trilogues. Each thematic session pairs one government representative, one civil society or academic voice, and one industry participant in moderated dialogue — preventing any single actor from dominating while ensuring all perspectives are heard in real time. This trilogue format has proven effective in EU legislative processes and is directly transferable. Pre-Dialogue open input sprints. The 30 April 2026 written-input deadline is a precedent worth institutionalising. Between annual sessions, rolling 60-day open consultation windows — in all six UN languages — allow communities that cannot travel to Geneva or New York to shape the agenda substantively. Scenario-based deliberation. Presenting delegates with concrete AI governance dilemmas — an AI grading system producing discriminatory outcomes, a generative AI tool reshaping curriculum in a low-income country — and deliberating on responses grounds abstract principles in actionable decisions. University-hosted intersessional hubs. Distributing thematic preparatory sessions across regional universities — including digital institutions like German UDS — decentralises the Dialogue physically and symbolically, embeds academic expertise directly into the process, and develops local governance capacity simultaneously. This would make higher education a structural partner, not a peripheral participant.

Five examples stand out as models the Dialogue should draw upon and scale. EU AI Act (2024) - risk-tiered regulation. The EU's landmark legislation classifies AI systems by risk level, banning unacceptable-risk applications and imposing transparency and conformity obligations on high-risk systems. Its explicit prohibition of AI-based social scoring and real-time biometric surveillance in public spaces sets a precedent for human rights-based governance. Its limitation - EU-only jurisdiction - is precisely the gap the Dialogue can address through a regulatory interoperability mechanism. UNESCO AI Competency Frameworks (2024) - operationalised principles. Unlike generic AI ethics declarations, UNESCO's frameworks for teachers and students translate principles into 12 progressive competencies with implementation guidance. They are the most actionable educational AI governance instrument available - but remain non-binding. The Dialogue should formally endorse and resource them as the global baseline for AI literacy in education. OECD AI Principles - the first intergovernmental standard. Endorsed by 46+ countries and the G20, the OECD Principles have directly shaped the EU AI Act and multiple national frameworks. Their five-pillar structure - inclusive growth, human-centred values, transparency, robustness, accountability - provides the conceptual architecture the Dialogue can build upon rather than duplicate. ITU AI Skills Coalition (2025) - multilateral capacity delivery. Launched at Davos with 25+ partners including AWS, Microsoft, and governments, it targets democratised AI training globally. This public-private model of capacity delivery is scalable and already operational - the Dialogue should designate it as the primary implementation vehicle for the AI capacity-building workstream. Z-Inspection methodology - trustworthiness assessment in practice. Developed for high-stakes AI deployment, Z-Inspection provides an interdisciplinary process for assessing AI trustworthiness in context - including in higher education. Its pilot in generative AI for universities demonstrates that sector-specific safety assessment is feasible. It offers the Dialogue a ready-made methodology for educational AI certification.