Digital Transformations for Health Lab (DTH-Lab)

A successful Global Dialogue on AI Governance would move beyond high-level principles to actionable, implementable outcomes. Key indicators of success include: - Agreement on practical governance pathways, particularly hybrid models that combine state regulation, industry implementation and independent oversight - Clear recognition of youth as a distinct stakeholder group with commitments to integrate developmental and well-being considerations into AI governance - Progress toward operationalising human rights frameworks, including child rights, into enforceable AI governance mechanisms - Identification of global minimum standards for safety, transparency and accountability, particularly in high-impact sectors such as health - Strengthened international coordination reducing fragmentation across governance approaches - Commitments to address AI risks effectively Importantly, the Dialogue should result in concrete follow-up mechanisms ensuring that discussions translate into sustained collaboration, policy development and implementation.

• Safe, secure and trustworthy AI
• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

These priorities reflect the need to address AI governance in a holistic and context-sensitive manner, particularly in relation to health systems and youth well-being. The social, economic, ethical and technical implications of AI are central as AI is not only a technological tool but an active influence on adolescent development, social relationships and mental health. While AI offers opportunities to improve access to health information, learning and support, it also introduces risks such as overreliance, emotional attachment to AI systems and reduced real-world social engagement. Ensuring safe, secure and trustworthy AI requires governance approaches that move beyond isolated models. Hybrid governance frameworks - combining legally binding safeguards, technical implementation and independent oversight - offer a more balanced approach to managing risks while enabling innovation. The protection and promotion of human rights is essential, particularly for children and adolescents who face distinct vulnerabilities. Existing ethical frameworks have not been sufficiently translated into enforceable protections, resulting in gaps in data protection. Finally, transparency, accountability and human oversight are critical to ensuring that AI systems can be effectively monitored and governed. Current approaches often lack clear mechanisms for auditing, oversight and redress, especially in high-impact sectors such as health. Strengthening these mechanisms is essential to building trust. This submission draws on insights from Digital Transformations for Health Lab (DTH-Lab)'s forthcoming AI in Health paper series (2026) which examines the intersection of AI governance, youth well-being and health systems. The series will be available in May 2026: https://www.dthlab.org/resources

While the identified themes capture key dimensions of AI governance, several cross-cutting issues require more explicit attention. First, the developmental and life-course impacts of AI, particularly on children and adolescents, are not sufficiently reflected. AI is not only shaping access to information and services but also influencing identity formation, social relationships and mental well-being. These effects require a more explicit, developmentally informed framing. Second, there is a need to better recognise AI as a social and behavioural influence, rather than solely a technical or economic tool. Emerging risks, such as emotional reliance on AI systems, parasocial interactions and behavioural manipulation, do not fit neatly within existing categories but have significant implications for well-being. Third, the tension between commercial incentives and public interest outcomes remains underexplored. Many AI systems are designed to maximise engagement and monetisation which may conflict with user well-being, particularly for younger populations. Fourth, data representativeness and "data poverty", especially in health contexts, require greater emphasis. Gaps in data quality and diversity can lead to exclusion, bias and reduced effectiveness of AI systems, disproportionately affecting vulnerable populations. Finally, there is a need to focus more explicitly on the implementation gap between principles and practice. While human rights and safety are recognised, mechanisms to operationalise these commitments across different contexts remain limited. Addressing these cross-cutting issues would strengthen the Dialogue's ability to respond to the full societal impact of AI.

In the health sector, governance gaps are creating both risks and missed opportunities. Key challenges include: - Lack of adolescent-specific protections despite unique vulnerabilities - Insufficient regulation of AI-driven health tools including mental health applications - Limited accountability mechanisms for harms such as misinformation, emotional manipulation and privacy violations - Persistent data gaps leading to biased or unrepresentative AI systems At the same time, AI presents opportunities to: - Improve access to health information and services particularly for underserved populations - Support early intervention and personalised care - Enhance health system efficiency and resilience However, these benefits will not be realised without governance frameworks that ensure equity, safety and trust. Strengthening governance capacity is essential to both mitigate risks and unlock these opportunities.

The AI Dialogue can play a critical role in advancing international cooperation by: - Facilitating alignment across governance approaches reducing fragmentation - Supporting the development of shared standards and principles that are actionable and context-sensitive - Providing a platform for multi-stakeholder engagement including underrepresented groups - Promoting knowledge exchange particularly on emerging risks and best practices Importantly, the Dialogue can help shift focus from principles to implementation, supporting the translation of commitments into practical governance mechanisms.

The AI Dialogue should build on and connect with existing international and multi-stakeholder initiatives that have advanced work on AI governance across sectors. These include efforts by the OECD on AI in health, WHO guidance on the use of AI in healthcare and established international human rights frameworks including those relating to children and young people. Multi-stakeholder initiatives that bring together governments, industry, civil society and academia are also critical for developing practical and context-sensitive approaches. The added value of the AI Dialogue lies in its ability to connect these efforts, reduce fragmentation and promote coherence across governance frameworks. It can serve as a platform to align principles with implementation ensuring that existing guidance translates into actionable and context-specific governance mechanisms. The Dialogue can also benefit from engaging with applied research and policy initiatives, including work that integrates youth perspectives and health system considerations such as DTH-Lab's work on AI governance in health. By linking global frameworks with practical implementation and diverse stakeholder perspectives, the Dialogue can strengthen coordination, avoid duplication and accelerate progress toward responsible and inclusive AI governance.

Stakeholders can contribute by sharing evidence and practical insights, engaging in dialogue across sectors and supporting implementation efforts within their respective contexts. To enable this, the Dialogue should prioritise: - inclusive participation across regions, sectors and disciplines - opportunities for underrepresented groups, particularly young people, to engage in a meaningful and structured way In particular, ensuring youth engagement is critical. Young people are among the most affected by AI systems, especially in areas such as health and well-being, yet remain underrepresented in governance discussions. Their inclusion should go beyond symbolic participation and involve structured opportunities to contribute to discussions and decision-making processes. The overall structure of the Dialogue should support both broad participation and practical, outcome-oriented exchanges, enabling diverse stakeholders to contribute while advancing shared understanding and actionable insights.

Young people, particularly adolescents, are significantly underrepresented in AI governance discussions, despite being major users and affected groups. Other underrepresented groups include: - populations from low- and middle-income countries - marginalised communities affected by data gaps Inclusion can be strengthened through: -structured youth engagement mechanisms -targeted outreach and capacity-building Ensuring participation by diverse stakeholders is essential for equitable and effective AI governance.

1. Hybrid governance models which bring together legally binding regulation, industry standards and independent oversight offer a balanced approach to managing risks while enabling innovation. These models allow governments to set enforceable safeguards, while industry develops technical solutions and civil society contributes accountability and public interest perspectives. 2. Risk-based frameworks, particularly those adapted to sector-specific contexts such as health, provide a practical way to prioritise oversight. By focusing on levels of risk and potential harm, these frameworks enable more proportionate and targeted governance, especially in high-impact areas such as clinical decision-making and mental health applications. 3. Safety-by-design approaches are particularly important for youth-facing technologies. Embedding safeguards into system design such as age-appropriate features, limits on engagement-driven design and protections against harmful content or interactions can help mitigate risks before they occur, rather than relying solely on post-hoc regulation. 4. The integration of human rights principles into AI governance is also critical. This includes aligning systems with existing frameworks, such as child rights principles and ensuring safeguards against privacy violations, bias, exclusion and emotional harm. However, these principles must be translated into enforceable requirements, including transparency obligations, independent audits and clear accountability mechanisms. Taken together, these approaches highlight the importance of governance that is context-specific, enforceable and responsive to evolving risks, particularly for vulnerable populations such as young people.