MyData Global

A successful first Dialogue must do more than produce declarations. It must deliver three foundational outcomes. First, it must establish a shared, operational understanding of human-centric AI governance — one in which individuals, including children, are active agents, not merely subjects. Digital systems must be co-designed with human wellbeing at the centre, protecting data, empowering autonomy, and enabling people to thrive. Second, it must produce concrete commitments on inclusion and equity. Countries with less AI capacity have had limited ability to shape international debates that will directly affect their populations. A successful Dialogue begins closing this asymmetry by co-designing governance with — not for — underrepresented regions and communities. Third, it must initiate a structural link between AI governance and binding human rights frameworks, including the Convention on the Rights of the Child. Children's data is increasingly used to train AI systems without adequate oversight or meaningful consent. This is not a technical anomaly — it is a rights deficit that AI amplifies. Success is not a communiqué. It is a set of interoperable commitments, accountability mechanisms, and inclusive processes that states and civil society can act upon between sessions.

• Safe, secure and trustworthy AI
• Interoperability of governance approaches
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

1. Safe, Secure and Trustworthy AI Children's data is increasingly collected by edtech, gaming, and social media platforms to train AI systems - often without meaningful consent, adequate transparency, or child-rights impact assessments. For MyData4Children, safe and trustworthy AI is not a technical standard alone; it is a developmental and rights imperative. AI systems interacting with children must be governed by strict ethical standards, particularly when processing data related to emotional states, preferences, or behaviour. Chatbots and conversational AI tools must be scrutinised for their psychological influence - tone, language patterns, and personality cues can shape children's social development in ways that are difficult to audit or reverse. Accountability must be embedded throughout the entire technical stack delivering child-facing services, not only at the platform level. If one element fails, the whole system becomes unsafe. A global baseline of unacceptable AI uses - including systems that exploit children's vulnerabilities or distort their behaviour - must be established and enforceable across jurisdictions. 2. Interoperability of Governance Approaches Children do not experience data risks within a single jurisdiction. Their digital lives span platforms, services, and borders simultaneously, yet the governance frameworks meant to protect them remain fragmented, sector-siloed, and often mutually incompatible. For MyData4Children, interoperability is not merely a technical requirement - it is the condition under which children's rights can be consistently upheld regardless of where a service is based or which regulatory regime applies. Children and their guardians must be able to move between services without forfeiting data rights or protections. Open standards, common data governance practices, and interoperable conformity assessment mechanisms are essential to prevent regulatory arbitrage that leaves children unprotected in the gaps between systems. A modular, scalable reference architecture - adaptable to both high-capacity and lower-capacity regulatory environments - ensures that interoperability serves equity, not only efficiency. Without it, the strongest national frameworks simply displace risk rather than reduce it. 3. Protection and promotion of human rights Human rights is the foundational priority. The current system is not working for children or families. In the physical world, the protection of children is more straightforward; in the digital world, parents are often powerless. Children's data and digital experiences lie with service providers, not with the children or their guardians. This structural imbalance is a rights deficit that AI governance must address explicitly. 4. Transparency, accountability, and human oversight These are indispensable complements. Children and their circle of trust - parents, guardians, educators - must be able to understand how decisions are made using children's data, receive alerts on risks, and challenge unfair or opaque processes. The entire data value chain must be auditable and accountable.

Two cross-cutting issues require dedicated attention beyond the current thematic framework. Children's data as a distinct governance category. Children experience the world holistically - blending the physical, digital, social, and emotional. Fragmenting these experiences with isolated or inconsistent policies creates more risk than solutions. Age-generic frameworks fail to account for children's evolving capacities and developmental stages. The Dialogue must treat children's data governance as a standalone pillar, not a subcategory of general data protection. Children's data in AI training. Children's personal data is increasingly used to train AI systems - often without their knowledge or meaningful consent. This raises profound ethical concerns, particularly regarding purpose limitation, data minimisation, and informed consent. Governance must go beyond legal compliance to ensure children's developmental vulnerabilities are not exploited at scale. Additionally, digital literacy and data civics must be treated as cross-cutting lenses applied to all thematic discussions. Governance mechanisms that do not build children's and families' capacity to understand and act on their data rights risk rendering all other protections structurally ineffective. Protecting children online requires ethical digital transformation - not just legal compliance.

The governance gaps most acutely affecting children and families can be characterised across three dimensions. Fragmented accountability. The digital services ecosystem — spanning platforms, infrastructure providers, and content intermediaries — lacks clarity in how responsibilities are shared. Edtech, social media, and gaming platforms collect detailed behavioural and interest-based data from children. This process is often opaque, serves primarily commercial outcomes, and is rarely auditable by children or their guardians. If one element of the technical stack fails, the entire system becomes unsafe. This is particularly challenging for SMEs that rely on third-party partnerships. Asymmetrical power over children's data. Children frequently accumulate multiple digital identities created by adults — through school accounts, parental setups, or third-party platforms. This fragmentation undermines accountability and coherent safeguarding. The data is rarely accessible to those it concerns most. Opportunity — convergence of frameworks. The convergence of the EU AI Act, the Convention on the Rights of the Child, UNICEF's Data Governance Fit for Children programme, 5Rights Foundation Age Appropriate Design, and MyData principles creates a policy alignment moment. The Dialogue can leverage this to drive global minimum standards for AI governance that are child-rights-centred from the outset — not retrofitted after the fact

The AI Dialogue occupies a unique institutional position: neither a negotiating forum nor a purely technical advisory body, but a deliberative space at the intersection of science, policy, and multi-stakeholder practice. For MyData Global and MyData4Children, its most significant contribution to cooperation is to function as a norm convergence mechanism — not producing binding law, but facilitating compatible governance architectures grounded in shared values. Concretely, the Dialogue can advance cooperation by treating interoperability as a human rights enabler. Human-centric data governance requires that individuals — including children — can move between services and jurisdictions without forfeiting their rights. Open standards and common practices must be used to ensure data portability and user freedom across borders. The Dialogue can also create peer-learning mechanisms on children's AI governance — structured exchanges among national regulators, civil society, and technical bodies on what age-appropriate AI design, data minimisation for minors, and child-rights impact assessments look like in practice. Finally, the Independent International Scientific Panel on AI must engage key stakeholders and communities of practice — including children, parents and guardians, their circle of trust, child rights organisations and human-centric data operators — to ensure its assessments reflect lived realities, not only laboratory or corporate environments

The Dialogue should not begin from a blank institutional canvas. A substantial ecosystem of governance frameworks and multi-stakeholder mechanisms already exists and must be explicitly connected to the Dialogue's work. 1) MyData Declaration and Operator Framework. The MyData Declaration articulates an approach for a human-centric approach to data sharing which have been operationalised, eg., across health, education, and mobility sectors globally and provide a directly applicable architecture for AI governance contexts. 2) MyData4Children. Building on both the MyData Declaration and UNICEF's child rights frameworks, MyData4Children has adapted these principles specifically for children's data contexts, co-designing approaches with families, educators, policymakers, and technologists. Its work on children-centric design, consent, accessibility, and accountability provides a ready-made reference framework for the Dialogue. 3) 5Rights Foundation and the Age Appropriate Design Code. The UK Children's Code — developed with the active involvement of 5Rights Foundation — is the first statutory code of practice setting out specific data protections for children, requiring digital services to offer high levels of privacy protection by default. Critically, 5Rights Foundation is now working with IEEE to develop a suite of standards for age appropriate digital services, based on principles that include presenting information in an age-appropriate way, upholding children's rights, and putting the child ahead of commercial interests. Policymakers globally are already looking to adopt the Code's underlying principles in their own legislation. The Dialogue should build on this momentum by recognising age-appropriate design as a global governance standard for AI systems that interact with or affect children. 5rights + 2 4) UNICEF Data Governance Fit for Children (DG4C). The DG4C framework translates the needs and interests of children into actionable, child-rights-centred outcomes and constitutes the most developed international framework for child-specific AI governance currently available. It should be formally referenced by the Dialogue. The Dialogue's added value is integration, translation, and accountability — linking these existing commitments to measurable implementation, not restating principles already established.

The credibility of the AI Dialogue depends on the genuine integration of diverse stakeholders throughout its design — not only in side events or observer roles. Civil society and communities of practice should be embedded in thematic working groups, with structured speaking rights and co-authorship of session summaries. MyData Global, with members and hubs across six continents, can model how networked civil society contributes substantively to governance deliberations. Children and young people require dedicated participation mechanisms informed by child rights principles — not tokenistic youth panels. Accountability must rest with digital service providers, not with children or guardians navigating unsafe-by-design systems. Private sector participants should engage through a transparency-first framework, required to disclose governance practices relevant to the thematic agenda rather than limiting contribution to promotional statements. Format recommendations: • Structured pre-Dialogue consultations with documented synthesis, disaggregated by stakeholder category. • Dedicated tracks for under-resourced participants, including translation support and asynchronous engagement options. • A standing civil society advisory body with rotating regional representation, providing continuity between annual sessions. • Publication of all submitted evidence and session records in accessible formats and multiple UN languages within 30 days of each meeting

The communities most affected by AI governance failures are systematically the least present in international deliberations. The Dialogue must address this through structural design, not aspirational language. Children and young people are the most consequential underrepresented group. Children are not merely a vulnerable group — they are agents of the future digital society. Their right to participate in decision-making is frequently overlooked, and age-generic policies fail to account for their evolving capacities. Dedicated youth tracks, age-appropriate formats, and partnerships with child rights organisations are essential. Communities in the Global South face dual exclusion: as primary recipients of AI-enabled surveillance and economic extraction, and as marginal participants in governance conversations. The Dialogue must allocate material resources — not only speaking slots — to ensure substantive participation. Guardians, educators, and circles of trust are critical but invisible in most governance processes. Legal approaches must empower guardians, not merely shift responsibility onto them. Practical measures: dedicated fellowships for civil society from low-income countries; pre-Dialogue community consultations with results formally incorporated into session agendas; and investment in data literacy that enables communities to articulate their governance needs on their own terms.

The standard UN conference model — high-level statements followed by panel discussions — is insufficient for the complexity and urgency of AI governance. The Dialogue should pilot formats that generate substantive outputs and enable genuine deliberation. Co-design workshops with structured outputs. Rather than panels where stakeholders present pre-formed positions, working sessions should task mixed groups — governments, civil society, technical experts, and affected communities — with producing joint governance artefacts: model clauses, design principles, or gap analyses. MyData Global has demonstrated this approach in collaboration with OECD, co-hosting workshops where participants developed strategies for embedding ethical data use in government policies and public spending. Child-rights impact simulations. For the children's data and human rights themes, structured simulations involving child advocates, educators, and technologists can surface governance gaps that expert presentations alone would not identify. MyData4Children has sustained dialogues with UNICEF, national ministries, and technology companies that provide a ready network for such formats. Living documentation. Session outputs should be published as evolving documents with clear version control, enabling stakeholders to track how their contributions influenced subsequent discussions. Asynchronous and multilingual participation pathways should complement physical sessions to expand geographic reach.

Several existing approaches provide concrete models the Dialogue can learn from and build upon. MyData Operator Model. The MyData approach establishes that people - including children and their guardians - have control over their data; that multiple providers of infrastructure-level services exist; and that services are interoperable and mutually substitutable. This architecture has been operationalised across health, education, and mobility sectors, offering a replicable template for human-centric AI governance. MyData4Children Principles in Practice. The six MyData principles - human-centric control, integration, empowerment, portability, transparency, and interoperability - have been adapted specifically for children's data contexts. Published on mydata.org, this framework provides actionable guidance for designing AI systems and data governance structures that genuinely serve children's best interests, accounting for the full ecosystem around them including families, educators, and service providers. EU AI Act risk-based framework. As set out in MyData4Children's submission to the UN AI Dialogue, Articles 5, 9, 10, and 14 of the EU AI Act provide a model for globally interoperable minimum standards. Mandatory child rights impact assessments, extended risk management requirements, and age-appropriate transparency obligations should be integrated into these standards. Protecting children online - and governing AI in their interests - requires ethical digital transformation, not just legal compliance.