IFBPWA

A successful first Global Dialogue on AI Governance should be measured by whether it establishes clear, actionable direction that countries and institutions can realistically implement. For me, the primary marker of success is alignment. Not uniformity, but a shared understanding of core principles like transparency, accountability, safety, and fairness as the baseline for how AI is designed, deployed, and governed across borders. Without this anchor, we invite fragmented governance, regulatory arbitrage, and growing systemic risk. Equally critical is inclusion. Global AI conversations are still dominated by a small cluster of governments and firms, often from the most technologically advanced economies. A meaningful dialogue must deliberately centre smaller states, developing economies, and underrepresented communities including women and those most affected by AI-driven decisions, so that governance reflects the realities of all, not just the loudest few. I also believe the dialogue must deliver practical pathways to implementation. High-level ethics language is no longer sufficient. We need concrete outputs such as; model policy guidance, implementation toolkits, sector-specific guardrails, and cooperative mechanisms that can be adapted in contexts as diverse as public administration, financial services, education, and health. Finally, success requires ongoing accountability and continuity. AI governance cannot be treated as a single event or communiqué. The Global Dialogue should establish a standing structure for follow-up: regular review, shared benchmarks, and transparent reporting on progress and gaps. If this first convening can move the global community from conversation to coordination, it will do more than open a discussion—it will lay a credible foundation for responsible AI that serves people in every region, not just those already at the table.

• Transparency, accountability, and human oversight
• Protection and promotion of human rights
• Safe, secure and trustworthy AI
• AI capacity-building

From my perspective, all four areas are critical, but the most urgent priorities for action and engagement are AI capacity-building and transparency, accountability, and human oversight, supported by a strong foundation in safe, secure, and trustworthy AI. First, AI capacity-building is essential. There is a growing gap between those who understand and can use AI effectively and those who cannot. Without intentional investment in education and access, particularly for women, small nations, and underserved communities we risk creating a system where opportunity is unevenly distributed. Capacity-building ensures people are not just users of AI, but informed participants who can question, adapt, and lead. Second, transparency, accountability, and human oversight must be prioritized. AI systems are already influencing decisions in finance, healthcare, and governance. If individuals do not understand how these systems work or cannot challenge their outcomes, we risk reinforcing bias and eroding trust. Practical frameworks that make AI more understandable and auditable are necessary to ensure responsible use. While safe, secure, and trustworthy AI is foundational, it must be approached in a way that is practical and enforceable, not just theoretical. Trust is built through consistent, visible safeguards. Finally, the protection and promotion of human rights must remain central. However, this requires moving beyond high-level commitments to ensuring that AI systems reflect the realities of diverse populations and do not unintentionally exclude or harm them. Overall, my priority is to ensure that AI governance is not only discussed at a global level, but implemented in ways that are inclusive, practical, and accessible.

In my view, the cross-cutting issues need to be made more explicit: institutional capacity for implementation, gender-responsive governance, and the political economy of AI. First, there is a gap between AI capacity-building and the institutional capability to govern AI day to day. Many public institutions, especially in small and developing states, do not yet have the legal, procurement, supervisory and audit capabilities required to translate principles into enforceable practice. Without investment in this "last mile" of governance capacity people, processes, and oversight interoperable frameworks will remain largely aspirational. Gender and intersectionality need to be treated as a structural design requirement, not only as part of the "social and ethical implications" theme. Women and marginalized communities are often the first to experience AI-related harms in employment, migration, social protection and finance, yet they are still underrepresented in governance fora and technical decision-making. A successful Dialogue should explicitly address gender-responsive impact assessment, participation, and redress as a cross-cutting obligation across all themes. Lastly, the political economy of AI-who funds, owns, and controls critical infrastructure, models, and data-cuts across open-source debates, interoperability, and human rights. Concentrated power over compute, platforms and standards can undermine inclusive governance, even where principles are well-articulated. The Dialogue should therefore consider safeguards around dependency, access, and accountability in AI value chains, so that smaller nations and communities are not permanently locked into rules they did not meaningfully shape.

In my context—small island, service-driven, and deeply connected to global systems—the gaps in AI governance are already visible in three places: public services, financial regulation, and everyday digital life. On the public side, there is strong interest in using AI for efficiency, but institutional capacity is uneven. Connectivity, data quality, and legacy systems still limit what can realistically be deployed, and most agencies do not yet have clear standards for procurement, risk assessment, or redress when automated decisions go wrong. This creates a real risk of importing "off-the-shelf" systems whose assumptions do not fit our legal, cultural, or demographic reality especially for migration, social protection, and education, where the stakes are highest. In financial services, we are at a regulatory inflection point. Supervisors are beginning to articulate expectations around board accountability, model validation, and transparency for AI use, but market participants range from highly sophisticated global firms to very small local entities. The challenge is to build a risk-based framework that is credible internationally, without overwhelming smaller players or hard-coding bias into high-risk use cases such as credit. The opportunity is that, as a small jurisdiction, we can move quickly. We can design interoperable, principle-based rules, pilot them in focused sectors, and create governance models that other small states can adapt—particularly around fairness, explainability, and human oversight in low-data, high-impact environments. If we get this right, AI in our region will not just be something done to us, but an area where we help shape what responsible, human-centred implementation looks like for small and vulnerable economies.

The AI Dialogue can become the central public infrastructure for international cooperation on AI governance, if it is designed for alignment, implementation, and accountability—not just discussion. First, it can create a structured space to align priorities and language across regions and stakeholder groups. By clustering debates around shared goals—such as managing risks, distributing benefits, and aligning rules—the Dialogue can help translate a crowded landscape of principles and initiatives into a more coherent, interoperable roadmap. Second, it can serve as a bridge between evidence and policy. Working in tandem with the Women In AI Bermuda on AI, the Dialogue can ensure that emerging technical insights on risks, capabilities, and mitigation feed directly into multilateral discussions and eventual norm-setting. Third, it offers a unique opportunity to institutionalize inclusion. Properly designed formats and processes can give smaller states, developing economies, civil society, and underrepresented communities structured access—not just speaking slots—into global rule-shaping. This is essential if AI governance is to reflect diverse realities rather than only the perspectives of a few large actors. Finally, the Dialogue can drive follow-through, by anchoring a practical, time-bound roadmap for cooperation. Clear milestones, transparent reporting, and continuity across annual cycles can help move the system from ad hoc consultations to predictable, rules-informed collaboration. In short, the AI Dialogue can turn scattered efforts into a shared architecture one that makes it easier for countries and sectors, including small and vulnerable ones, to move in the same direction on AI governance while retaining their specific contexts.

In Bermuda, the AI Dialogue should connect with three main strands of work that are already underway: public-sector AI policy, financial-sector regulation, and broader digital/fintech strategy. On the public side, Bermuda has adopted a Government AI Policy that embeds human-in-the-loop decision-making, alignment with PIPA and PATI, and strong transparency and auditability requirements for AI used in public services. This is being implemented through pilots under the wider Digital Transformation Programme, which is modernising cloud, data and service design capabilities across government. The Dialogue could add value by documenting Bermuda's experience as a small jurisdiction operationalising human rights and accountability requirements in real government workflows, and by offering a platform to test interoperable standards that other small states can adapt. In financial services, the Bermuda Monetary Authority has issued a discussion paper and subsequent guidance signalling a principles-led, proportionate approach to AI governance—focused on board accountability, risk assessment, model validation and transparency, embedded within existing prudential and conduct frameworks. This sits alongside a broader fintech strategy and digital asset framework that already position Bermuda as a regulated innovation hub. Here, the AI Dialogue could help translate Bermuda's "sandbox to system" experience into global good practice for small but globally integrated financial centres, and ensure that emerging international norms reflect the realities of proportionate supervision. Across these areas, the added value of the AI Dialogue would be to: - Create structured channels for Bermuda to both contribute to and draw from global standards; - Elevate small-state perspectives on capacity, proportionality and inclusion; and - Support practical cooperation (e.g., shared toolkits, peer learning) that accelerates safe, human-centred AI deployment at our scale.

The AI Dialogue will only be credible if it is designed for shared work. Member States should use the Dialogue to align on principles, mandates and incentives that can travel across borders—anchoring AI governance in human rights, development priorities, and concrete obligations on transparency and accountability. Regulators and standard‑setters can translate these into interoperable supervisory expectations, technical benchmarks, and procurement criteria that make responsible AI the default rather than the exception. Private sector actors—especially those controlling compute, models and data—should be held to specific, time‑bound commitments on disclosure, risk management and support for capacity-building in less-resourced contexts. Civil society, academia and affected communities should shape problem definition, impact assessment, and redress models, not just comment on end products. To enable this, I would recommend a structured format that combines: Regular multistakeholder plenary sessions for agenda‑setting and political signalling. Thematic working groups tasked with producing short, practical outputs such as checklists, model clauses, and implementation toolkits. Regional and virtual consultations to surface context‑specific risks and opportunities, particularly from small states and the Global South. A clear, published calendar and reporting cycle, so stakeholders can prepare substantively and track progress over time. In short, the Dialogue's structure should make it easy for each stakeholder group to do what only they can do, while keeping all of us accountable to a shared, global roadmap.

Several voices remain structurally underrepresented in global AI governance debates: small and vulnerable states, women and girls, particularly in the Global South and Caribbean—frontline public servants, and communities already subject to heavy datafication (migrants, low‑income workers, informal sector, persons with disabilities). These are often the people and institutions who experience AI first as surveillance, automated exclusion or opaque scoring, not as innovation. Yet their lived experience rarely shapes the design of safeguards, standards or accountability mechanisms. The AI Dialogue can change this by hard‑wiring inclusion into its design: Set explicit balance targets for participation by region, gender, and stakeholder type in all formal sessions and working groups, and report publicly against them. Resource participation through travel support, interpretation, and accessible virtual formats for small missions, civil society, and community-led organisations that otherwise cannot be in the room. Create dedicated hearing sessions on high‑risk domains (e.g. migration, social protection, labour, political participation) where affected communities, not only experts, set the agenda and present evidence. Partner with regional and feminist networks to bring grounded, gender‑responsive and intersectional analysis into the core of the Dialogue, rather than treating it as a side event. If the Dialogue can move underrepresented groups from "consulted" to "co‑designers," it will be better equipped to govern AI in the world we actually live in, not only in the models we build.

For the AI Dialogue to be genuinely useful, the engagement formats have to match the speed, complexity, and lived impact of AI—not just the formality of multilateral processes. I would highlight four types of innovative formats: 1. Use‑case labs, not just panel discussions Small, time‑boxed "labs" where governments, regulators, companies, and civil society work through a concrete scenario together—e.g. AI in border management, credit scoring, or social protection—and map risks, safeguards, and governance options in real time. These labs should be designed to produce short, practical artefacts (checklists, questions to ask vendors, model clauses) that can be reused. 2. Stakeholder "role‑reversal" sessions Structured dialogues where, for example, regulators are asked to argue from the perspective of small civil society groups, or large platforms have to present from the standpoint of a small island state with limited capacity. This helps surface blind spots and power asymmetries in a way that traditional statements often do not. 3. Intersessional community clinics Virtual "clinics" in between annual meetings, hosted in partnership with regional bodies and networks, where underrepresented stakeholders can bring specific challenges for peer feedback and documentation into the Dialogue. 4. Open drafting and feedback windows Where draft guidance, roadmaps, or principles are posted on an accessible platform with structured opportunities for written, public comment from all stakeholder groups—including those who cannot attend in person. A commitment to show how feedback shaped the final output is essential for trust. Taken together, these formats would shift the Dialogue from a series of prepared interventions to a working space where people co‑design the tools, norms, and guardrails that AI governance actually needs.

Bermuda is already putting some of the core ideas of responsible AI governance into practice, even without a standalone AI law. The AI Dialogue could draw on Bermuda's experience to illustrate: - How human-in-the-loop and transparency requirements can be implemented in practice; - How proportional, risk-based supervision of AI in finance can work; and - How small states can link AI governance to broader digital and fintech strategies without losing sight of rights and inclusion.