My Futures Lab

A successful first Global Dialogue on AI Governance would move beyond discussion and deliver clear signals of collective direction. First, it should establish a shared global baseline for safe and trustworthy AI, grounded in risk-based governance, human rights, and meaningful human oversight. Without convergence on minimum standards, fragmentation will persist and trust will erode. Second, it should secure agreement on a core set of prohibited AI practices, including systems that exploit human vulnerabilities or manipulate behaviour—particularly in relation to children. If governance cannot define its red lines, it cannot command legitimacy. Third, success requires a concrete commitment to interoperability across regulatory approaches. This includes alignment on risk classification, transparency requirements, and certification mechanisms, enabling cooperation without imposing uniformity. Fourth, the Dialogue must deliver tangible progress on inclusion. This means scaling capacity-building, supporting Global South participation in standard-setting, and ensuring that governance frameworks do not become barriers to entry. Fifth, it should recognise that data governance is foundational to AI governance, and initiate steps toward shared standards on data quality, bias mitigation, and privacy-preserving cross-border collaboration. Finally, a successful outcome would embed a child-centred, rights-based approach, aligned with the Convention on the Rights of the Child, as a test of system integrity. AI systems that are safe for children are more likely to be safe for society as a whole. In sum, success will be measured not by the breadth of participation, but by whether the Dialogue establishes a credible path from principles to coordinated global action.

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

First, advancing safe, secure and trustworthy AI is foundational. Without a shared baseline of safety, risk management, and system integrity, trust in AI will remain uneven and fragile across jurisdictions. Second, interoperability of governance approaches is essential to avoid regulatory fragmentation. As AI systems operate across borders, alignment on core principles-such as risk classification, assurance mechanisms, and compliance frameworks-is critical to enable cooperation while respecting national contexts. Third, the protection and promotion of human rights must remain central. AI systems increasingly shape access to services, information, and opportunity. Anchoring governance in international frameworks, including the Convention on the Rights of the Child, ensures that technological development does not come at the expense of fundamental rights-particularly for vulnerable groups. Fourth, transparency, accountability, and human oversight are necessary to operationalise these principles. Without clear lines of responsibility, explainability, and meaningful human control, governance frameworks risk becoming aspirational rather than enforceable.

Yes. While the listed themes are comprehensive, several cross-cutting and emerging issues merit more explicit attention to ensure AI governance is both effective and equitable. First, child-centred AI governance remains underrepresented. Children face distinct risks-such as manipulation, profiling, and long-term data exploitation-yet are rarely treated as a priority group. Embedding child rights, in line with the Convention on the Rights of the Child, as a cross-cutting principle would strengthen all thematic areas and provide a clear benchmark for trustworthy AI. Second, the political economy of AI requires greater focus. The concentration of power in a small number of firms and countries through control of data, compute, and talent risks entrenching global inequalities. Without addressing these structural dynamics, progress on capacity-building and inclusion may be limited. Third, compute and infrastructure governance is an emerging gap. Access to high-performance computing is increasingly essential for AI development, raising questions about equitable access, environmental sustainability, and technological dependence. Fourth, accountability and liability frameworks need further development. As AI systems become more autonomous, clearer mechanisms are required to assign responsibility, ensure redress, and maintain public trust. Finally, intergenerational impacts are insufficiently recognised. Decisions on AI design and governance today will shape economic, social, and rights outcomes for decades. Incorporating longer-term perspectives into governance processes is therefore critical. Addressing these issues would strengthen the Dialogue's ability to deliver inclusive, forward-looking, and globally balanced AI governance.

Governance gaps in safe, trustworthy, and interoperable AI are already shaping outcomes across sectors working with and for children, with both risks and opportunities emerging. A key challenge is fragmentation across regulatory approaches. Diverging standards increase compliance complexity and create uncertainty for organisations operating across borders, particularly those handling sensitive data relating to children. This risks uneven levels of protection and limits the scalability of responsible AI solutions. Second, insufficient integration of child rights into AI governance creates tangible risks. In sectors such as education, social services, and digital platforms, AI systems are being deployed without consistent safeguards against bias, profiling, or manipulation. The absence of systematic child rights impact assessments—aligned with the Convention on the Rights of the Child—means that harms may go unrecognised or unaddressed. Third, limited interoperability in accountability and assurance mechanisms constrains trust. Organisations face multiple, sometimes conflicting, requirements for transparency, documentation, and oversight, reducing efficiency and discouraging innovation—particularly for smaller actors. At the same time, significant opportunities are emerging. Growing alignment around risk-based governance models, including developments such as the EU AI Act, provides a foundation for convergence. There is also increasing recognition of the need for human oversight, transparency, and data quality standards, which can enhance trust if implemented coherently. Finally, the current moment creates an opportunity to embed child-centred safeguards from the outset, rather than retrofitting protections later. If aligned globally, this could position child rights as a cornerstone of trustworthy AI—benefiting not only vulnerable groups, but society as a whole. Addressing these gaps will determine whether AI governance evolves as a source of fragmentation or as a driver of shared, equitable progress.

First, it can establish a shared global baseline for AI governance. By convening Member States and stakeholders under the legitimacy of the United Nations, the Dialogue can drive convergence around core elements such as risk-based governance, human rights protections, and meaningful human oversight—while allowing flexibility in national implementation. Second, it can act as a platform for advancing interoperability between governance approaches. Rather than promoting a single model, the Dialogue can facilitate alignment on key building blocks—risk classification, transparency requirements, and assurance mechanisms—reducing fragmentation and enabling cross-border cooperation. Third, the Dialogue can strengthen inclusive participation, particularly by elevating the role of Global South countries in shaping norms and standards. This includes supporting capacity-building, sharing best practices, and ensuring that governance frameworks reflect diverse social, economic, and cultural contexts. Fourth, it can catalyse practical cooperation mechanisms, such as international regulatory sandboxes, shared testing environments, and collaboration on standards for data governance and AI assurance. These initiatives can translate high-level commitments into operational outcomes. Fifth, the Dialogue can reinforce the centrality of human rights in AI governance, aligning efforts with international frameworks such as the Convention on the Rights of the Child and ensuring that protections—particularly for vulnerable groups—are embedded across all areas of cooperation. Ultimately, the value of the AI Dialogue will lie in its ability to act not only as a forum for exchange, but as a driver of coordinated global action, bridging divides and shaping a more coherent, inclusive approach to AI governance.

The AI Dialogue should deliberately build on, and help connect, the growing but fragmented ecosystem of international AI governance initiatives to avoid duplication and strengthen coherence. Key foundations include the work of the Organisation for Economic Co-operation and Development on AI principles and measurement frameworks, the Group of Seven Hiroshima AI Process on international guiding principles and code of conduct, and the standards and technical coordination led by the International Organization for Standardization. In parallel, regional regulatory frameworks such as the EU AI Act are shaping risk-based governance approaches that are already influencing global practice. The Dialogue should also connect with multistakeholder and technical initiatives, including AI safety research networks, digital rights coalitions, and child online protection frameworks grounded in the Convention on the Rights of the Child. These efforts collectively reflect a growing but dispersed architecture of AI governance. The added value of the AI Dialogue lies in its unique legitimacy as a universal, UN-led convening space. Unlike existing initiatives, it can bring together all Member States on equal footing, alongside civil society, industry, and technical experts, to identify areas of convergence and gaps. Its key contribution should be to: Translate fragmented principles into shared global baselines Promote interoperability between regulatory and technical standards Support inclusive participation, particularly from underrepresented regions Enable coordination between existing initiatives rather than replacing them Most importantly, the AI Dialogue can serve as the bridge between normative discussions and operational cooperation—ensuring that AI governance evolves from parallel efforts into a coherent and inclusive global system.

Different stakeholders can contribute most effectively if the AI Dialogue is structured as a continuous, action-oriented multistakeholder process, rather than a series of isolated consultations. Member States should play the central role in defining convergence points on governance, particularly around risk-based approaches, interoperability, and baseline safeguards for safe and trustworthy AI. Their participation should focus on identifying areas where alignment is both necessary and feasible, rather than negotiating full harmonisation. The private sector and technical community should contribute evidence on system capabilities, limitations, and implementation realities, including emerging risks and best practices in safety, assurance, and data governance. Their input is essential for ensuring that governance frameworks are technically feasible and innovation-compatible. Civil society, including child rights organisations, should be enabled to provide structured input on rights impacts, societal risks, and equity considerations, ensuring that governance reflects lived experience and protects vulnerable groups in line with the Convention on the Rights of the Child. Academia should support the Dialogue through independent analysis, foresight, and evaluation of governance models, helping identify long-term implications and unintended consequences. In terms of structure, the AI Dialogue would benefit from: Thematic working tracks aligned with key governance pillars (safety, interoperability, rights, capacity) A standing multistakeholder expert group to synthesise inputs and identify convergence areas Regular policy-to-practice sessions linking principles with real-world implementation cases A mechanism for Global South participation and capacity support, ensuring equitable engagement The Dialogue should also prioritise outputs over process, including synthesis reports, shared principles, and identified pathways for cooperation. Ultimately, its effectiveness will depend on whether it becomes a bridge between fragmented initiatives and coordinated global action, rather than a purely consultative forum.

Several voices remain systematically underrepresented in global AI governance discussions, limiting the legitimacy and completeness of emerging frameworks. First, children and young people are rarely directly represented, despite being among the most affected by AI systems over time. Their interests are often mediated through adult institutions, which can dilute or overlook specific risks such as behavioural profiling, manipulation, and long-term data use. More direct inclusion—through youth advisory structures or dedicated child rights mechanisms aligned with the Convention on the Rights of the Child—would strengthen governance legitimacy. Second, stakeholders from the Global South, particularly regulators, civil society actors, and local innovators, remain underrepresented in standard-setting and technical discussions. This risks embedding priorities and assumptions from a narrow set of jurisdictions into global norms. Meaningful inclusion requires sustained capacity-building, funding for participation, and co-design of governance frameworks rather than late-stage consultation. To address these gaps, the AI Dialogue could introduce: Structured participation pathways for youth and child rights representatives Dedicated funding mechanisms for Global South participation in technical and policy fora Ultimately, inclusion must go beyond representation. It must ensure that these perspectives meaningfully shape standards, priorities, and governance outcomes, not merely observe them.

First, "policy-to-practice" working labs could bring together Member States, industry, civil society, and technical experts to test governance ideas against real-world AI use cases. These labs would help translate high-level principles into implementable approaches, particularly in areas such as risk classification, transparency, and assurance mechanisms. Second, scenario-based foresight sessions could be used to explore emerging risks and opportunities, including long-term societal impacts. These sessions would allow stakeholders to stress-test governance options against plausible future developments, improving resilience and anticipatory capacity. Third, a global regulatory sandbox network could be established under the Dialogue to support cross-border experimentation. This would enable countries—especially those with limited regulatory capacity—to pilot governance approaches in collaboration with peers and technical experts, reducing fragmentation while encouraging innovation. Fourth, the Dialogue could adopt structured youth and child rights engagement formats, including dedicated panels or advisory circles aligned with the Convention on the Rights of the Child. This would ensure that those most affected by long-term AI development trajectories have a meaningful voice in shaping governance. Finally, rotating regional engagement hubs could ensure broader participation, particularly from the Global South, by decentralising dialogue processes and reducing barriers to engagement. Together, these formats would shift the AI Dialogue from a consultative forum into a dynamic governance platform, capable of generating actionable outcomes, building trust, and accelerating international alignment on AI governance.

Several existing policies, practices, and platforms already offer concrete foundations for effective AI governance and can inform the work of the AI Dialogue. A key example is the EU AI Act, which introduces a risk-based regulatory framework distinguishing between unacceptable, high-risk, and lower-risk AI systems. Its requirements for risk management, data governance, transparency, and human oversight provide a structured model for operationalising trustworthy AI at scale. At the international level, the AI principles developed by the Organisation for Economic Co-operation and Development have helped establish widely recognised norms on transparency, robustness, and accountability. Similarly, the Group of Seven Hiroshima AI Process has advanced voluntary international guiding principles and a code of conduct for advanced AI systems, contributing to early convergence among leading economies. In parallel, regulatory sandbox approaches-used in several jurisdictions-offer a practical mechanism for testing AI systems in controlled environments, enabling innovation while managing risk. These approaches are particularly valuable for emerging economies seeking to build regulatory capacity without stifling development. From a rights-based perspective, frameworks anchored in the Convention on the Rights of the Child provide an essential benchmark for ensuring that AI systems account for the specific vulnerabilities and long-term interests of children. Collectively, these initiatives demonstrate that effective AI governance is already emerging in a fragmented form. The central challenge is not the absence of solutions, but the need to connect, align, and scale them globally. The AI Dialogue can play a key role in identifying convergence points and enabling interoperability across these diverse approaches.