Bahamas Government

uccess for the first Global Dialogue on AI Governance must be judged by one criterion: does the outcome serve the entirety of UN membership, or does it codify the preferences of nations already at the frontier of AI development? For Small Island Developing States, success requires three things. First, a concrete, funded commitment to fair and equitable access to AI, not access on terms dictated by dominant technology powers, but access that preserves every nation's right to determine how AI is developed and governed within its borders. Fair access is not charity; it is a structural requirement for meaningful participation in the global digital economy. Second, internationally recognized data protection principles that apply across borders, including clear norms governing how data generated by citizens of developing nations is collected, processed, and monetized by entities outside their jurisdiction. Data is the raw material of AI. SIDS are net exporters of it, with almost no share of the returns. Third, formal recognition that open-source AI models, open data, and open infrastructure are governance instruments, not merely technical preferences. For nations without the resources to build sovereign AI systems, open-source frameworks are the only realistic pathway to technological self-determination. Success is not a declaration. It is a framework with mechanisms, timelines, and accountability, one that SIDS helped write, not one they were asked to endorse after the fact.

• Open-source software, open data and open AI models
• AI capacity-building
• Protection and promotion of human rights
• Social, economic, ethical, cultural, linguistic and technical implications of AI

Synthesized four strategic priorities for island nations' AI governanceThese four priorities reflect the specific vulnerabilities and strategic interests of Small Island Developing States at this pivotal moment in AI governance. AI capacity-building is the prerequisite for everything else. SIDS cannot be meaningful participants in global AI governance as regulators, innovators, or negotiating partners without domestic technical expertise, legislative frameworks, and institutional infrastructure. Capacity-building must be treated as a structural investment, not a philanthropic afterthought, with binding commitments and measurable delivery timelines. Open-source software, open data, and open AI models represent the most viable pathway to technological sovereignty for small states. Proprietary AI systems built and controlled by a small number of corporations create permanent dependency. Open-source frameworks allow SIDS to build, adapt, and govern AI systems that reflect their own languages, cultures, legal traditions, and developmental priorities. The Dialogue should affirm open-source as a public good and resist governance frameworks that inadvertently privilege closed, proprietary systems. Protection and promotion of human rights encompasses the data protection dimension that is critically underserved in current AI discourse. For SIDS, this is not abstract. It is the daily reality of citizens whose data is continuously extracted by platforms and AI systems with no accountability to local law, no consent architecture, and no benefit-sharing mechanism. Human rights in the AI context must explicitly include the right to data protection, algorithmic accountability, and freedom from discriminatory automated systems. Social, economic, ethical, cultural, and linguistic implications speaks to the outsized displacement risk AI poses to service-sector and tourism-dependent economies, and to the cultural erasure risk embedded in AI systems that do not reflect the diversity of the world's languages, identities, and knowledge systems.

Two issues demand explicit recognition that the current thematic framework does not adequately capture. Fair access to AI as a development right. The existing themes address capacity-building and implications, but stop short of naming the structural access inequality that defines the AI landscape for SIDS. The concentration of AI development, compute infrastructure, frontier models, proprietary datasets, and talent, in a tiny number of countries and corporations is not a market outcome to be managed at the margins. It is a governance failure that will deepen global inequality if left unaddressed. The Dialogue must establish fair access to AI as a foundational principle: nations have a right to participate in the benefits of AI development commensurate with their contribution to the data ecosystems that make AI possible. Cross-border data governance as the infrastructure layer of AI. AI governance cannot be meaningfully separated from data governance, yet the two remain largely siloed in multilateral discourse. For SIDS, this gap is particularly damaging. Our citizens' behavioral, health, financial, and cultural data flows continuously to platforms and AI training pipelines domiciled abroad, subject to the laws of other jurisdictions, generating value that never returns to its origin. The Dialogue should advance binding norms on cross-border data flows, data benefit-sharing, and the extraterritorial obligations of AI developers, recognizing that data sovereignty is not protectionism but a legitimate developmental interest. Together, these two issues form the structural foundation beneath the listed themes. Without addressing access inequality and cross-border data governance, commitments on safety, rights, and capacity risk becoming aspirational language that leaves the existing power asymmetry intact. SIDS are not asking for special treatment. We are asking for a governance architecture that reflects the world as it actually is.

For Small Island Developing States, AI governance gaps are not theoretical concerns. They are present-tense economic and sovereign risks. The most significant challenge is structural exclusion from the AI value chain. SIDS collectively generate substantial volumes of tourism, health, financial, and cultural data that feeds global AI systems. Yet the infrastructure to process that data, the models trained on it, and the commercial value derived from it exist entirely outside our jurisdictions. We contribute to AI development without participating in its governance or sharing in its returns. This is not an emerging risk. It is already the operating reality. The capacity gap compounds this. Most SIDS lack the legislative frameworks, regulatory institutions, and technical talent to evaluate, audit, or negotiate with AI systems and their developers on equal terms. When a large platform deploys an AI-driven content moderation, credit-scoring, or hiring system in our markets, we have limited recourse and almost no visibility into how those systems work or whom they disadvantage. The linguistic and cultural dimension is equally urgent. AI systems trained predominantly on English-language, Western data perform poorly for our populations, distort our cultural representations, and create a slow erosion of linguistic diversity that no single governance framework has yet seriously addressed. The opportunities are real but conditional. AI holds genuine potential for SIDS in climate resilience modeling, public service delivery, healthcare access, and economic diversification. However, realizing that potential requires open-source tools we can adapt, data protection frameworks we can enforce, and capacity-building investments that are sustained rather than episodic. The governance gaps do not merely slow our progress. They determine whether SIDS are participants in the AI future or subjects of it.

The AI Dialogue occupies a unique position in the international governance landscape precisely because it is convened under the authority of the General Assembly, the only multilateral body where every nation, regardless of size or economic weight, holds equal standing. That legitimacy is the Dialogue's most valuable asset, and it must be used deliberately. The Dialogue can advance international cooperation in three distinct ways. First, it can establish a shared normative floor. Existing AI governance initiatives are fragmented across regional bodies, bilateral agreements, and voluntary frameworks, most of which were designed by and for technologically advanced economies. The Dialogue can develop baseline principles, on fair access, data protection, and open infrastructure, that apply universally and that smaller nations can reference in their own legislative and regulatory development. Second, it can function as a coordination layer between existing mechanisms. The Dialogue does not need to replace the OECD AI Principles, the EU AI Act, the African Union Data Policy Framework, or CARICOM's emerging digital governance work. It can map these frameworks, identify where they conflict or leave gaps, and establish interoperability norms that allow nations operating across multiple regulatory environments to do so without being forced to choose between them. Third, and most importantly for SIDS, it can give developing nations a structured, recurring seat at the table. Not as observers or beneficiaries, but as co-authors of the rules that will govern technology affecting their populations. The Dialogue should establish formal mechanisms for SIDS and LDC participation that go beyond representation and confer genuine agenda-setting influence. The value of the Dialogue is proportional to whose priorities it reflects. Cooperation that does not include the full UN membership is coordination among the powerful, not governance for the world.

The AI Dialogue should not begin from a blank page. A significant body of work already exists, and the Dialogue's added value lies in connecting, elevating, and universalizing it rather than duplicating it. Key initiatives the Dialogue should build upon include the OECD AI Principles and the Global Partnership on AI, which provide technically grounded frameworks but reflect a membership weighted toward high-income economies. The UNESCO Recommendation on the Ethics of AI is the most universally endorsed normative instrument to date and should serve as a foundational reference. The ITU's AI for Good platform and the work of the UN Secretary-General's AI Advisory Body offer multilateral entry points with existing SIDS engagement. At the regional level, the African Union Data Policy Framework and emerging CARICOM digital governance discussions represent governance innovation from the Global South that deserves amplification, not marginalization. The Dialogue's distinct added value is threefold. It carries General Assembly authority, giving any principles or frameworks it produces a legitimacy that voluntary or regional instruments cannot match. This matters enormously for small states seeking to reference international norms in domestic legislation or in negotiations with large technology platforms. It can operationalize the link between AI governance and development finance. Capacity-building commitments made in governance forums too often remain unfunded. The Dialogue, connected to UN development architecture, is positioned to attach resources to obligations in ways that other AI governance bodies are not. Finally, it can center the voices that existing initiatives have structurally underrepresented. SIDS, landlocked developing countries, and least developed countries are not niche constituencies. They represent a majority of UN member states. A Dialogue that reflects their priorities would not be a concession to the margins. It would be the most representative AI governance exercise ever undertaken.

The AI Dialogue will only produce durable governance outcomes if its structure reflects the diversity of those governed by AI, not merely those who build or regulate it. This requires deliberate design, not default formats inherited from other multilateral processes.Governments must remain the primary decision-making actors, but their deliberations should be formally informed by three additional constituencies. Civil society organizations, particularly those working at the intersection of digital rights and development, bring ground-level accountability and community impact perspectives that no government delegation can fully represent. The technical community, including open-source developers, academic researchers, and standards bodies, provides the expertise necessary to ensure that governance frameworks are technically coherent and not inadvertently captured by incumbent industry interests. The private sector has a legitimate role, but its participation should be structured to prevent regulatory capture, with clear disclosure requirements and separation between advisory input and decision-making authority.For SIDS specifically, the format must address the participation barrier that cost and capacity create. Many small island nations cannot sustain permanent delegations in New York or Geneva. The Dialogue should establish hybrid participation as a permanent feature, not a pandemic-era concession, with dedicated support for SIDS delegates including translation, technical briefings, and pre-session capacity preparation.The structure should include a formal preparatory process that allows developing nations to arrive with consolidated positions rather than reacting to agendas set by others. Regional consultative mechanisms, including through CARICOM, the Pacific Islands Forum, and AOSIS, should feed directly into the Dialogue's working documents.A single annual session is insufficient for a technology that evolves faster than any governance cycle. The Dialogue should operate on a continuous model with thematic working groups that report to plenary, ensuring momentum between formal sessions.

The voices most consequentially absent from global AI governance discussions are, with few exceptions, the voices of those most exposed to AI's risks and least positioned to shape its development. Small Island Developing States and Least Developed Countries are the most structurally underrepresented governments in current AI governance forums. Most existing initiatives, from the OECD AI Principles to the G7 Hiroshima Process, were designed within and for high-income economies. SIDS and LDCs are often consulted after frameworks are drafted rather than included in their design. The Dialogue must create dedicated representation mechanisms, not token seats, that give these nations substantive agenda-setting influence. Within countries, several communities are systematically absent. Indigenous peoples, whose knowledge systems, languages, and cultural data are increasingly incorporated into AI training without consent or attribution, have no formal standing in any major AI governance body. Women and girls, who face disproportionate exposure to AI-enabled discrimination, surveillance, and online harm, are underrepresented both as technical contributors and as policy voices. Youth, who will live longest with the consequences of decisions made now, are treated as beneficiaries of AI governance rather than as participants in it. Linguistically, the dominance of English in AI governance discourse excludes the majority of the world's population from meaningful engagement. Working documents, consultation processes, and deliberative sessions must be available in all UN official languages as a baseline, with active outreach into non-official language communities. Inclusion cannot be achieved through open-door policies alone. It requires proactive investment: travel and participation support for delegates from small and developing states, interpretation infrastructure, accessible formats for persons with disabilities, and dedicated consultation processes for communities that do not engage through formal governmental channels.

The default format of multilateral governance, formal plenary statements followed by negotiated text, is poorly suited to a technology that evolves faster than any drafting cycle and that most delegates experience primarily as users rather than as technologists. The AI Dialogue needs formats that build genuine understanding, not just formal representation. Structured scenario exercises should be incorporated into the preparatory process. Rather than asking delegations to negotiate abstract principles, the Dialogue should present concrete governance scenarios, an AI system deployed in a small island nation's healthcare system without local data protection safeguards, a large language model trained on indigenous cultural data without consent, a predictive policing tool exported from one jurisdiction to another, and ask delegations to work through the governance response. This produces more grounded deliberation and surfaces the specific gaps that matter most to different constituencies. Rotating regional host sessions would distribute the Dialogue's center of gravity beyond New York and Geneva. Holding thematic sessions in Bridgetown, Suva, Nairobi, or Dhaka would not only reduce the participation burden on developing nations but would signal that global AI governance is genuinely global. Persistent digital deliberation platforms, designed for structured policy input rather than social media dynamics, would allow civil society, technical experts, and community representatives to contribute between formal sessions. These platforms should be multilingual by design and moderated to prevent capture by well-resourced actors who can flood consultation processes with volume. Finally, the Dialogue should integrate direct testimony from affected communities as a formal agenda item, not a side event. Hearing from a Pacific Islander whose traditional ecological knowledge has been scraped into an AI training dataset, or a Caribbean worker whose employment was displaced by an algorithmic hiring system, is not theater. It is the evidence base that governance decisions should rest on.

Several policy approaches and governance models offer concrete lessons for the AI Dialogue, particularly for Small Island Developing States navigating AI governance with limited institutional capacity. Open-source digital governance architecture demonstrates that small nations can build sovereign, interoperable public infrastructure without dependence on a single vendor or foreign system. Where governments have adopted open-source data exchange and digital identity frameworks, they have created replicable models for AI-ready public infrastructure that other small states can adapt to their own legal and developmental contexts. The principle that small states can be governance innovators, not only governance recipients, is the most transferable lesson the Dialogue can carry forward. Data governance frameworks that treat data as a developmental resource, rather than simply a regulatory compliance matter, offer a more useful foundation for SIDS than purely technical standards. Provisions addressing cross-border data flows, benefit-sharing, and community data rights speak directly to the extraction dynamic that currently disadvantages developing nations in the AI value chain. The UNESCO Recommendation on the Ethics of AI, as the most universally endorsed AI governance instrument to date, provides a normative baseline the Dialogue should build from rather than duplicate. Its inclusion of cultural diversity, environmental sustainability, and the needs of vulnerable populations reflects a broader conception of AI governance than most technical frameworks offer. Open-source AI research initiatives that prioritize multilingual model development and treat linguistic and cultural diversity as design requirements rather than afterthoughts demonstrate that frontier AI development is possible outside the proprietary model. These initiatives should be formally recognized and supported within the Dialogue's framework. The Dialogue should document, connect, and scale approaches that center developing-world priorities rather than defaulting to frameworks produced by the wealthiest AI-producing nations.