University of Surrey; Women in AI

A successful first Global Dialogue on AI Governance should deliver concrete, actionable outcomes that move beyond discussion and into coordinated implementation. First, it should establish a shared baseline for AI governance. This includes agreement on core principles such as safety, accountability, transparency, and human oversight, with practical guidance on how these principles translate into real systems across different contexts. Second, it should define a clear framework for international interoperability. Governments and institutions are moving at different speeds, but fragmented approaches will limit impact. The Dialogue should produce alignment mechanisms that allow national policies, standards, and regulatory approaches to work together without creating barriers to innovation or deployment. Third, it should prioritise implementation pathways. This means identifying practical models for applying governance in areas such as public services, education, and industry, supported by case studies and tested frameworks. A strong outcome would include guidance that organisations can adopt immediately, not just high-level recommendations. Fourth, it should embed inclusion and capability development. Many regions and institutions lack the infrastructure and expertise to engage effectively with AI governance. The Dialogue should define strategies to build capacity, particularly for youth, educators, and emerging economies, ensuring broad participation in the AI ecosystem. Finally, it should create continuity. A clear roadmap, working groups, and measurable milestones are essential to maintain momentum beyond the first Dialogue. Success will be defined by whether the Dialogue produces shared direction, practical tools, and a sustained global mechanism for coordination and accountability in AI governance.

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

My priorities focus on ensuring that AI governance is practical, inclusive, and capable of being implemented at scale. AI capacity-building is critical. There is a widening gap between those who can design, deploy, and govern AI systems and those who cannot. Without targeted investment in skills, education, and institutional capability, governance frameworks will remain theoretical. Priority should be given to building capacity among youth, educators, public sector institutions, and emerging economies. Interoperability of governance approaches is equally urgent. Different national and organisational frameworks are developing rapidly, but without alignment they risk fragmentation. Establishing mechanisms that enable policies, standards, and regulatory approaches to work together is essential to support both innovation and responsible deployment. Transparency, accountability, and human oversight are central to trust. Organisations need clear, practical guidance on how to implement these principles within real systems, including decision-making processes, data use, and automated workflows. This is where governance must move from principles to operational models. Finally, the broader social, economic, ethical, and cultural implications of AI must be addressed in a structured way. AI is already reshaping labour markets, education systems, and access to opportunity. Governance must actively consider these impacts to ensure equitable outcomes and avoid reinforcing existing inequalities. Together, these priorities reflect a need to move from high-level principles to coordinated, real-world implementation.

Several cross-cutting issues require urgent attention beyond the listed themes. First, the rise of AI-native systems and autonomous agents. Governance frameworks are still largely designed for static models, not systems that act, learn, and interact across environments. There is a need for standards that address multi-agent coordination, system-level accountability, and continuous monitoring. Second, the gap between policy and implementation. Many organisations lack practical guidance on how to translate governance principles into operational systems. There is a need for implementation frameworks, audit mechanisms, and measurable benchmarks that can be applied across sectors. Third, economic concentration and infrastructure dependency. A small number of providers control key components of the AI stack, including compute, models, and data infrastructure. This creates systemic risk and limits equitable access. Governance must address resilience, competition, and access to foundational resources. Fourth, workforce transition and organisational readiness. AI is not only a technical shift but an operational one. Many institutions are unprepared to redesign roles, processes, and decision-making structures. Governance should include guidance on organisational transformation, not only compliance. Finally, evaluation and verification. There is no consistent global approach to assessing AI system performance, safety, and societal impact. Standardised evaluation frameworks are needed to enable comparability, trust, and accountability across jurisdictions. These issues cut across all themes and are critical to ensuring that AI governance is effective, enforceable, and aligned with real-world system behaviour.

Governance gaps are already affecting the UK and the higher education and youth innovation sector in visible ways. The most significant challenge is the gap between policy ambition and operational readiness. Institutions are expected to adopt responsible AI, yet there is limited practical guidance on how to implement governance within real systems. This leads to inconsistent approaches across universities and organisations, slowing adoption and increasing risk. A second challenge is fragmentation. Different frameworks, institutional policies, and emerging standards are developing in parallel without clear alignment. This creates uncertainty for organisations working across sectors and limits interoperability, particularly for collaborative research, startups, and public sector partnerships. Capacity is another critical issue. There is a shortage of individuals with the skills to design, deploy, and govern AI systems effectively. This is particularly visible in youth education and early-stage innovation ecosystems, where demand for AI capability is growing faster than supply. At the same time, there are strong opportunities. The UK has a well-developed research base, active innovation ecosystems, and increasing engagement with AI governance. This creates a foundation to lead in developing practical, standards-aligned implementation models. There is also an opportunity to integrate governance directly into education and innovation programmes. By embedding principles such as transparency, accountability, and human oversight into how AI is taught and applied, it is possible to build a generation that is not only technically capable but also governance-aware. Addressing these gaps can position the UK as a leader in translating AI governance from policy into scalable, real-world practice.

The AI Dialogue can play a practical coordination role by turning fragmented efforts into aligned, implementable action. First, it can establish a common operating baseline. Not just principles, but shared definitions, minimum governance requirements, and reference models that countries and organisations can adopt without starting from scratch. Second, it can enable interoperability. The Dialogue should act as a bridge between different national policies, standards bodies, and regulatory approaches, ensuring they can work together rather than compete or conflict. This is critical for cross-border systems, research collaboration, and global markets. Third, it can accelerate implementation. The Dialogue can curate and validate real-world use cases, governance models, and tools that have been tested across sectors. This allows participants to move from discussion to adoption, reducing duplication and increasing speed. Fourth, it can support capacity-building at scale. By connecting governments, academia, and industry, the Dialogue can facilitate knowledge transfer, training models, and institutional capability development, particularly for regions with limited resources. Fifth, it can create continuity and accountability. Structured working groups, clear milestones, and ongoing reporting mechanisms are needed to ensure that outcomes are sustained beyond the event itself. The value of the AI Dialogue will be defined by its ability to move from high-level alignment to coordinated, cross-border implementation that is practical, measurable, and inclusive.

The AI Dialogue should build on existing standards, policy, and technical initiatives rather than duplicate them. Key foundations include the ISO/IEC JTC 1/SC 42 work on AI standards, the OECD AI Principles, and regulatory frameworks such as the EU AI Act. It should also connect with multi-stakeholder platforms like the Global Partnership on AI, as well as technical ecosystems driving open innovation. In addition, there are growing contributions from academia, innovation programmes, and industry-led collaborations that are already testing governance in practice. These provide valuable real-world insight but remain fragmented and often disconnected from policy and standards development. The added value of the AI Dialogue lies in integration and translation. First, it can act as a coordination layer, aligning outputs from standards bodies, regulators, and industry initiatives into a coherent global direction. This reduces duplication and helps ensure interoperability across jurisdictions. Second, it can bridge the gap between policy and implementation by identifying and scaling proven governance models from real-world deployments. This includes practical frameworks, evaluation methods, and operational guidelines that organisations can adopt. Third, it can create structured pathways for emerging stakeholders, including universities, startups, and youth-focused innovation ecosystems, to contribute to global governance discussions. Finally, it can provide continuity by linking these initiatives into an ongoing process with shared milestones and accountability. Its value will come from connecting what already exists into a system that is aligned, actionable, and globally relevant.

Different stakeholders should contribute based on their role in designing, deploying, and governing AI systems, with a structure that ensures practical outputs. Governments should provide policy direction, regulatory frameworks, and national priorities, while committing to interoperability and shared baselines. Standards bodies such as International Organization for Standardization and International Electrotechnical Commission should translate these into implementable standards and technical guidance. Industry should contribute real-world deployment experience, including use cases, risk management practices, and operational challenges. Academia should provide research, evaluation methods, and independent validation. Civil society should ensure that human rights, inclusion, and societal impacts are fully represented. Startups and innovation ecosystems should bring insight into emerging applications and constraints faced by smaller actors. The Dialogue should be structured in three layers. First, thematic working groups aligned to priority areas, producing draft outputs in advance of the main convening. Second, a central plenary that focuses on alignment, decision-making, and endorsement of key recommendations, rather than open-ended discussion. Third, an implementation track that captures tested frameworks, case studies, and tools, creating a shared repository that participants can apply in practice. To ensure continuity, the Dialogue should establish ongoing working groups with clear milestones, reporting cycles, and accountability mechanisms. This structure allows broad participation while ensuring that contributions are translated into coordinated, practical outcomes.

Several perspectives remain underrepresented in global AI governance, and their absence limits the practicality and inclusiveness of current frameworks. First, youth and early-career practitioners. They are among the most affected by AI-driven changes in education and employment, yet rarely participate in governance discussions. Inclusion requires structured youth advisory panels, integration into formal working groups, and dedicated pathways from education and innovation programmes into policy forums. Second, educators and institutional implementers. Universities, schools, and training providers are responsible for deploying AI in learning environments and building future capability, but their operational insight is often missing. They should be included through sector-specific working groups and direct representation in governance design processes. Third, startups and early-stage innovators. Much governance input comes from large organisations, while smaller actors face different constraints in resources, compliance, and deployment. Inclusion requires targeted engagement mechanisms, simplified contribution processes, and representation through innovation ecosystems and accelerators. Fourth, practitioners from emerging and under-resourced regions. Many governance frameworks are shaped by a limited set of geographies. Expanding participation requires funded access, regional consultation hubs, and capacity-building programmes that enable meaningful contribution rather than symbolic inclusion. Finally, interdisciplinary perspectives are often fragmented. Social scientists, behavioural experts, and community practitioners need to be integrated alongside technical experts to ensure governance reflects real-world impact. Inclusion should be built into the structure of the Dialogue, not treated as an add-on. This means formal representation, accessible participation formats, and clear mechanisms for translating diverse input into decision-making and implementation.

Traditional panel discussions will not be enough. Engagement formats should be designed to produce decisions, not just dialogue. First, structured problem-solving labs. Small, cross-sector groups work on a defined governance challenge such as model accountability or cross-border data use, with clear outputs required within a fixed timeframe. Each group delivers a short, actionable framework or set of recommendations. Second, implementation showcases. Instead of abstract presentations, organisations present tested governance models, including what worked, what failed, and measurable outcomes. This grounds discussions in real systems and accelerates adoption. Third, policy-to-practice translation sessions. Mixed groups of policymakers, practitioners, and technical experts take existing principles or regulations and map them into operational steps, identifying gaps and constraints. Fourth, scenario-based simulations. Participants engage in structured simulations of AI-related risks or governance failures, such as system misuse or cross-border incidents, to test coordination, response mechanisms, and accountability structures. Fifth, rapid consultation rounds. Short, focused sessions where underrepresented stakeholders, including youth, educators, and startups, provide direct input on specific questions, with their contributions formally captured and fed into outputs. Finally, continuous working groups supported by digital collaboration platforms. Engagement should not end with the event. Participants should be able to refine outputs, share data, and track progress over time. These formats shift the Dialogue from passive discussion to active co-creation, ensuring that engagement produces practical, implementable outcomes.

Several existing approaches provide practical models for effective AI governance. The EU AI Act is a strong example of a risk-based regulatory framework. It classifies AI systems by risk level and links this directly to obligations, creating clarity for organisations on compliance and accountability. Its strength lies in moving beyond principles to enforceable requirements. The OECD AI Principles provide a widely adopted foundation for responsible AI, particularly around transparency, accountability, and human-centred values. While high-level, they have influenced national policies and organisational frameworks globally. From a standards perspective, the International Organization for Standardization and International Electrotechnical Commission work through committees such as ISO/IEC JTC 1/SC 42 offers structured guidance on AI system lifecycle management, risk assessment, and governance processes. These are critical for operationalising governance within organisations. In practice, model governance frameworks used in industry, including internal audit processes, model documentation, and lifecycle monitoring, are increasingly important. These approaches provide concrete mechanisms for implementing transparency, accountability, and oversight in deployed systems. There are also emerging practices in education and innovation ecosystems, where governance principles are embedded directly into programme design. This includes integrating ethical review, transparency requirements, and human oversight into how AI systems are developed and tested by students and early-stage ventures. Together, these examples show that effective AI governance requires a combination of regulation, standards, and practical implementation models that organisations can adopt and adapt across different contexts.