Small Islands, Big Decisions

Abstract

Island states and island jurisdictions are no longer absent from artificial-intelligence policy. Iceland, Malta, Cyprus, Mauritius, the Dominican Republic, Singapore, Ireland, Aotearoa New Zealand and Sri Lanka have adopted national strategies or formally approved strategic programmes. Hawaiʻi has public-service guidance; the Faroe Islands have received formal ethical recommendations; and policy work is under way in Jamaica, the Isle of Man, Maldives and Trinidad and Tobago. This is substantial activity, but it does not amount to an island-centred approach. Most instruments reproduce a familiar national policy repertoire: innovation, investment, skills, public-service efficiency, data access, safety and international competitiveness. Their location on islands seldom changes the questions being asked.

This working paper argues that AI policy made on an island should not automatically be classified as island AI policy. An island-centred policy must address archipelagic inequalities, small-language and cultural continuity, community and Indigenous authority over data, dependence on overseas platforms and cloud services, cable and energy resilience, locally significant sectors, and the capacity to refuse unsuitable systems. Comparison of adopted strategies and emerging policy processes reveals important precedents—especially Icelandic language infrastructure, Singaporean attention to resource-constrained compute, Mauritian links between AI and resilience, and Caribbean proposals for regional pooling—but also persistent omissions. The paper proposes an Island-Centred AI Policy Test and introduces the AI-slands Island AI Policy Tracker as a transparent, updateable research resource. The decisive issue is not whether islands adopt AI quickly. It is whether island societies retain the practical authority to decide what AI is for, what it may use, and when dependence has become too costly.

Keywords: artificial intelligence; islands; public policy; small states; digital sovereignty; infrastructure; Indigenous data governance; resilience

Evidence note

The references are limited to sources whose existence and bibliographical details were checked against publisher, proceedings, government, intergovernmental or institutional records. Substantive claims were checked against complete texts or the relevant pages wherever these were accessible. DOI, repository and stable institutional links are supplied so that readers can inspect the evidence. Policy status is reported cautiously: an adopted strategy, operational guidance, a recommendation, a readiness assessment and a policy under development are not treated as equivalent. The inventory is a verified starting collection, not a claim to exhaustive global coverage.

1. From adoption to authority

National discussions of artificial intelligence often begin with a race. Governments are encouraged to adopt quickly, attract investment, train workers, release data, build compute and avoid being left behind. This language is especially persuasive in small jurisdictions. Limited administrative scale can be presented as agility; small markets can be recast as test beds; and concentrated public institutions may appear able to coordinate policy more rapidly than larger states. Yet the same conditions also create vulnerability. A government with a small technical workforce may depend heavily on consultants and vendors. A geographically dispersed archipelago may digitise public services while leaving outer islands with fragile connectivity. A small language may be officially protected but practically absent from the models through which citizens increasingly obtain information.

The first AI-slands working paper described insular AI sovereignty as the capacity of island communities to shape, govern and sometimes refuse the systems that affect them (Jennings, 2026a). The second applied that argument to culture and language, showing why inclusion in a model is not the same as authority over cultural data or representation (Jennings, 2026b). This paper turns to policy. It asks a simple question: when does an AI strategy made on an island become genuinely island-centred?

The distinction is necessary because geography alone proves little. Singapore's strategy is one of the world's most developed small-state AI programmes, but it is designed primarily around a densely connected global city and hub economy. Ireland and Aotearoa New Zealand are island states, yet their national strategies are not structured around islandness. Malta and Cyprus invoke small-state opportunity more readily than internal archipelagic inequality. Conversely, a document that is not labelled an AI strategy may be highly relevant to island governance: Caribbean proposals for shared digital infrastructure, for example, confront limited capacity and vendor lock-in directly (UNDP et al., 2024).

Island-centred policy is therefore not a separate ethical code added after a conventional innovation strategy. It is a way of identifying what conventional strategies make difficult to see: unequal connection between islands, the physical systems beneath supposedly immaterial services, the value of small languages and collective knowledge, and the difference between purchasing a capability and controlling it.

2. The emerging island policy landscape

The AI-slands policy inventory identified a first group of adopted national or jurisdiction-wide strategies. Iceland adopted a national AI policy in 2021 and subsequently connected implementation to a 2024–2026 action programme. Malta launched Malta: The Ultimate AI Launchpad in 2019. Cyprus approved a national AI strategy in 2020. Mauritius issued an early strategy in 2018 and launched a new National AI Strategy and FAIR Guidelines in April 2026. The Dominican Republic adopted its national strategy in 2023. Singapore launched strategies in 2019 and 2023 and published a substantial update in May 2026. Ireland refreshed its 2021 strategy in 2024; Aotearoa New Zealand published Investing with Confidence in July 2025; and Sri Lanka's Cabinet approved a national AI strategy following consultation (Government of Iceland, 2021; Government of Malta, 2019; Republic of Cyprus, 2020; UNDP, 2026; OECD.AI, 2026; Smart Nation Singapore, 2026; Government of Ireland, 2024; MBIE, 2025; Cabinet Office of Sri Lanka, 2024).

A second group consists of narrower instruments. The State of Hawaiʻi publishes data and AI guidance for state agencies, including guidance on generative assistants. In May 2026 the Faroese UNESCO Committee submitted recommendations on ethical AI to the minister responsible for children, education, research and culture. These called for a human-centred approach, public oversight, education and attention to artists' rights, but they were recommendations rather than adopted legislation or a final government strategy. Institutional policies add another layer but should not be mistaken for jurisdiction-wide policy (State of Hawaiʻi Office of Enterprise Technology Services, n.d.; Faroese Ministry of Children, Education, Research and Culture, 2026).

A third group is still developing policy. Jamaica's National AI Task Force published recommendations and a reconstituted task force is supporting further work. The Isle of Man has established a National AI Office to develop a strategy. Trinidad and Tobago is undertaking a national assessment, and UNESCO records the development of a Maldivian AI masterplan. A 2025 Pacific assessment reported that no Pacific Island country had then published a national AI strategy, while warning that the baseline could change as Fiji, Papua New Guinea and others developed initiatives (Jamaica National AI Task Force, 2025; Isle of Man Government, 2025; Ministry of Public Administration and Artificial Intelligence, 2026; UNESCO, 2025; Hua and Brimble, 2025). Absence claims must remain dated and qualified: policies may be newly released, unpublished, or poorly indexed.

The inventory demonstrates that islands are not simply waiting for metropolitan policy to arrive. It also exposes a problem of classification. Strategies, ethical frameworks, operational rules, consultation papers and readiness exercises are often reported together in international dashboards, although they create different obligations. A tracker that obscures these differences produces a misleading picture of preparedness. The formal existence of a strategy also says little about budgets, implementation, procurement practice, enforcement or community power.

3. Policies on islands are not necessarily island policies

Across the adopted strategies, five familiar priorities recur: economic growth, public-service modernisation, skills, trustworthy use and participation in international AI markets. All are legitimate. None is distinctively insular. What matters is what appears alongside them—and what does not.

3.1 Smallness is usually framed as agility

Malta and Singapore have used small scale as a strategic asset. Singapore's 2023 foreword calls it a “small island state” with large ambitions and builds a whole-of-nation programme around talent, industry, research, government, data, compute and a trusted environment (Smart Nation Singapore, 2023, pp. 3–5). This is much more than branding: it demonstrates that a small state can assemble institutional capacity and revise policy as the technology changes. The May 2026 update reports the creation of a National AI Council and refreshes ten enablers rather than pretending that a 2023 document remains sufficient (Smart Nation Singapore, 2026, pp. 3–6).

Yet the model cannot simply be exported to archipelagic or resource-constrained SIDS. Singapore is a global logistics, finance and data-centre hub with unusually strong state capacity. Its lessons concern purposeful coordination, sustained investment and policy revision, not a universal small-island development path. A strategy built around attracting global firms may deepen dependence elsewhere if the local state cannot negotiate contractual safeguards, retain staff or fund public alternatives.

3.2 Language is the strongest island-specific precedent

Iceland provides the clearest example of an island condition reshaping AI policy. Its language-technology programme brought universities, public institutions and private firms together to build corpora, speech tools, machine translation and writing support as open infrastructure (Nikulásdóttir et al., 2020, pp. 3414–3422). A later governmental account explains why a small commercial market cannot sustain these resources unaided and describes public funding, permissive licensing, community involvement and cultural evaluation (Ministry of Culture and Business Affairs et al., 2024, pp. 4–12).

This approach matters beyond Iceland. Global natural-language processing remains concentrated in a small minority of the world's languages (Joshi et al., 2020, pp. 6282–6293). A policy that promises citizens equal digital services while procurement assumes fluent performance in the dominant language may create a new tier of linguistic inequality. Language infrastructure should therefore be treated as a public capability, not merely as a cultural application to be funded after commercially attractive systems are procured.

The Icelandic case is not a complete model of island-centred governance. It does not by itself resolve internal geographic inequalities, Indigenous rights or dependence on global foundation models. It nevertheless demonstrates a decisive principle: the state may need to create and maintain resources that the market will not provide because cultural continuity has public value.

3.3 Infrastructure appears, but dependency is rarely governed

AI depends on electricity, water, data centres, chips, cloud services and telecommunications. The International Energy Agency estimates that a typical AI-focused data centre can use as much electricity as 100,000 households and stresses that local grid effects can be pronounced (IEA, 2025). This comparison is global rather than island-specific, but its implications for small electricity systems are evident. The relevant question is not only whether a jurisdiction can attract a data centre. It is which uses deserve scarce land, water and energy, who bears grid costs, and what happens when capacity is located abroad.

Singapore addresses this physicality unusually directly. Its 2023 strategy notes that compute requires power, carbon, water and land, all limited in Singapore, and proposes managed allocation and greener infrastructure (Smart Nation Singapore, 2023, pp. 49–50). The 2026 update places future compute within an “energy envelope”, land constraints and climate commitments, while combining local supercomputing with cloud and partner-centre access abroad (Smart Nation Singapore, 2026, pp. 30–32). This is a strong resource-governance precedent, even though the strategy does not analyse islandness as such.

For many islands, the more immediate risk is not excessive local compute but dependence on distant systems. Submarine-cable research shows why this matters. Franken et al. (2022, pp. 1–2, 8–13) found that many small and island jurisdictions had little route redundancy and documented the cascading social and economic effects of cable failure. Their 2020 network data are not a current audit, and national aggregation can hide internal inequalities, but the structural lesson endures: a digital public service is only as resilient as its weakest cable, landing station, power supply, satellite backup and repair arrangement.

Most AI strategies treat cloud services and foundation models principally as enablers. Island-centred policy must also treat them as dependencies. Public procurement should disclose where data and inference occur, whether a service functions during connectivity loss, how models can be changed or exited, whether records are portable, and which decisions remain possible locally when a vendor withdraws or changes price.

3.4 Data governance often begins with access rather than authority

National strategies commonly seek to unlock public and private data. Conventional safeguards—privacy, security, quality and lawful sharing—are essential, but they do not exhaust the issue. Some island data concern communities, customary knowledge, language, land and sea territories, or collectively held cultural resources. Individual consent and state ownership may not provide adequate authority for their reuse in AI.

The CARE Principles for Indigenous Data Governance place Collective Benefit, Authority to Control, Responsibility and Ethics alongside the technical emphasis of FAIR data (Carroll et al., 2020). They require attention to peoples, purposes and governance across the data lifecycle. They cannot be applied indiscriminately to every island community, and Indigenous sovereignty must not be reduced to a metaphor for local preference. Where Indigenous peoples are concerned, however, participation in a national consultation is not a substitute for their authority over Indigenous data.

Research on cultural-data partnerships reinforces this distinction. Spano and Zhang (2025, pp. 167–193) show how Indigenous data sovereignty can alter the design and oversight of public–private arrangements, including language technology. An island-centred strategy should therefore identify which data require collective protocols, how communities can place conditions on reuse, and whether meaningful refusal remains possible after digitisation.

3.5 Sectors are listed more often than island systems are understood

Several strategies mention agriculture, tourism, the ocean economy, climate services or disaster prediction. Mauritius's 2026 National AI Strategy and FAIR Guidelines are explicitly presented as locally grounded, inclusive and connected to sustainable development and resilience (UNDP, 2026). Cyprus includes resource efficiency, agriculture, weather forecasting and disaster prevention. Singapore's 2026 update treats port, air and land connectivity as a national AI mission (Smart Nation Singapore, 2026, pp. 11–12). These are useful beginnings.

An island-centred approach goes further by asking how sectors connect. Tourism demand affects housing, water, energy and seasonal labour. Fisheries policy joins ecological monitoring to customary access, food security and maritime jurisdiction. AI-assisted disaster response depends on local communications and trusted knowledge when central services fail. Inter-island transport determines whether a nominally national service is genuinely reachable. Policy should be organised around these relationships rather than a catalogue of technological use cases.

4. The Island-Centred AI Policy Test

The proposed test is not a league table and should not produce a single score that conceals political differences. It is a structured reading framework. Each dimension asks whether an instrument recognises a specifically island condition, creates a practical response, assigns responsibility, provides resources and permits public scrutiny.

4.1 Place before national averages

An island state may have high national connectivity while smaller islands experience slower links, fewer technicians and longer outage recovery. National adoption targets can therefore reward activity in the capital while widening the gap within the archipelago. Policies should publish geographically disaggregated measures, require offline and low-bandwidth service design, and assess whether centralisation removes work and judgment from outer islands.

4.2 Public capability, not permanent outsourcing

Small administrations cannot build every model or host every service. Sovereignty does not mean autarky. It means retaining enough legal, technical and institutional capability to choose suppliers, understand systems, protect data, negotiate collectively, maintain essential services and leave unsuitable arrangements. UNCTAD's SIDS digital-economy work recommends context-sensitive data governance, low-bandwidth and resilient public digital infrastructure, regional interoperability and shared tools where local capacity is limited (UNCTAD, 2022).

Regional cooperation is therefore important but not automatically benign. Evidence reviews of public-sector reform in SIDS show successful pooling in higher education, fisheries, auditing and trade, while also identifying conflicting interests, weak consultation and sovereignty concerns (Ismail, 2019). The right level of cooperation depends on the function. Shared testing laboratories, model evaluation, procurement clauses, incident reporting and specialist training may be more feasible than a single regional authority controlling all island data.

4.3 A right to human and community judgment

AI governance should preserve both individual redress and collective authority. The UNESCO Recommendation on the Ethics of Artificial Intelligence provides a global baseline across governance, data, environment, culture, education, work and health, while UNESCO and i4Policy argue that AI policy is too complex to be designed by a single stakeholder group (UNESCO, 2021; UNESCO and i4Policy, 2022). In islands, multistakeholder participation must not mean one capital-city workshop. It should include outer-island authorities, language communities, Indigenous governments and organisations, cultural institutions, small businesses, unions, civil society and people whose access is constrained.

Participation should continue after adoption. Communities need routes to identify errors, halt harmful uses and require review. A practical right to refuse may apply to a dataset, a high-risk automated decision, a culturally restricted use or the replacement of a face-to-face public service. It does not imply rejecting AI as a whole. It makes adoption conditional on continuing legitimacy.

5. Comparative lessons

Iceland: protect what the market will neglect

Iceland shows how language can become a strategic infrastructure commitment rather than a symbolic aspiration. Its transferable lesson is the combination of public investment, institutional coordination, open resources and speaker involvement. The unresolved question is how such programmes retain bargaining power when they cooperate with much larger model providers.

Singapore: plan for scale and physical limits

Singapore demonstrates unusually coherent state capacity, repeated policy revision and explicit planning for compute, land, energy and water. It also shows the limitations of the category “island state”: its hub-centred model may reveal little about dispersed archipelagos. The appropriate lesson is disciplined coordination and resource accounting, not imitation of a global-city growth strategy.

Mauritius and the Caribbean: link ethics to development capacity

Mauritius's new strategy and FAIR Guidelines connect responsible use with local innovation and resilience. The test will be whether these principles translate into budgets, procurement rules and public reporting. At the regional level, the SIDS 2.0 outcome statement proposes shared infrastructure, cybersecurity, pooled expertise, digital public goods and measures against vendor lock-in (UNDP et al., 2024, pp. 1–8). Its significance lies in treating limited capacity as a design condition rather than a temporary deficit.

Hawaiʻi and the Faroe Islands: jurisdiction and culture matter

Hawaiʻi's guidance shows the value of linking AI use to data classification, retention, privacy and quality inside public administration (State of Hawaiʻi Office of Enterprise Technology Services, n.d.). The Faroese recommendation places AI within the ministerial domain of culture, education and the Faroese language and calls for political oversight, public knowledge and artists' rights (Faroese Ministry of Children, Education, Research and Culture, 2026). Neither is a comprehensive island AI strategy, but both demonstrate that useful governance can develop below the level of a sovereign national strategy.

The Pacific: do not mistake a policy gap for an empty landscape

Where no dedicated strategy exists, international dashboards may imply that nothing is happening. This can erase digital-government work, customary governance, regional institutions and community technology practice. It may also invite imported templates that define readiness as resemblance to larger states. A Pacific approach should begin with regional and Indigenous institutions, disaster and cable resilience, linguistic plurality, ocean governance and the capacity of small administrations—not with a generic strategy awaiting localisation.

6. Ten commitments for an island-centred strategy

An island AI strategy should make ten commitments.

1. Map the archipelago. Publish subnational evidence on connectivity, public-service access, skills, benefits and harms rather than relying on national averages.
2. Treat language and culture as infrastructure. Fund datasets, tools, evaluation and long-term stewardship where markets cannot provide them.
3. Recognise collective authority. Build community and Indigenous data governance into law, procurement and system design.
4. Count physical costs. Assess energy, water, land, carbon, cable and repair implications before promoting compute-intensive development.
5. Plan for disconnection. Require essential services to have offline, low-bandwidth, backup and human-continuity arrangements.
6. Govern dependency. Disclose vendors, hosting, subcontractors and model changes; require portability, interoperability and credible exit plans.
7. Build local judgment. Invest in public-interest technical expertise, domain professionals, language specialists and independent scrutiny rather than permanent consultant dependence.
8. Start from island systems. Prioritise connected challenges in marine governance, fisheries, tourism, food, transport, health, education, demography and disasters.
9. Cooperate without disappearing. Pool specialist capacity regionally where useful while preserving transparent responsibility and the authority of participating jurisdictions and peoples.
10. Preserve refusal and redress. Ensure human review, accessible challenge, correction and the ability to stop uses that lack social or cultural legitimacy.

These commitments are deliberately institutional. Ethical principles matter, but a principle without a responsible body, resources, reporting and a remedy can coexist comfortably with business as usual.

7. The Island AI Policy Tracker

The accompanying AI-slands Island AI Policy Tracker converts this research into a living public resource. Each entry records the jurisdiction, instrument, date, policy status, principal themes, island-centred features, limitations, primary or authoritative source, and last verification date. Filters separate adopted strategies from guidance, recommendations and work in progress. The tracker avoids a single ranking because policy maturity and island-centredness are different variables: a technically sophisticated strategy may pay little attention to island conditions, while a short recommendation may articulate a strong cultural principle without implementation power.

The Tracker should be revised at least twice yearly and whenever an instrument changes status. Priority watch points include the developing policies of the Isle of Man, Jamaica, Maldives and Trinidad and Tobago; governmental responses to the Faroese recommendation; implementation reporting in early-adopter states; and new work in Pacific Island countries and subnational archipelagos. Corrections should preserve an audit trail so that later researchers can distinguish a changed policy from a corrected description.

The resource also establishes a research agenda. Future work can compare policy text with procurement contracts and public-service practice; measure whether benefits reach outer islands; document local responses to outages or vendor withdrawal; and examine whether language, cultural and Indigenous-data commitments alter actual model development. The Tracker is therefore not simply a catalogue. It is an empirical bridge between the AI-slands conceptual framework and comparative policy research.

8. Conclusion

Islands do have AI policies. The more difficult finding is that they rarely have policies in which islandness reorganises the analysis. Most strategies could be moved to a continental small state with relatively little alteration. They encourage innovation, skills, investment and responsible adoption, but seldom confront the combined realities of archipelagic inequality, cultural scale, collective data authority, fragile connections, constrained infrastructure and dependence on distant firms.

There are important precedents. Iceland treats a small language as public technological infrastructure. Singapore counts compute against land, energy, water and climate limits. Mauritius connects responsible AI to local needs and resilience. Caribbean cooperation proposals confront capacity and vendor lock-in. Hawaiʻi and the Faroe Islands show how operational and cultural governance can emerge within non-sovereign island jurisdictions. None provides a complete template, and no single template should be expected.

The aim is not island technological self-sufficiency. Islands have always developed through exchange, mobility and connection. The policy task is to ensure that connection does not remove the capacity to choose. AI becomes island-centred when island peoples and institutions can decide which problems deserve computation, which knowledge may be used, which dependencies are acceptable, and which relationships must remain governed locally. The measure of success is not adoption at any price. It is durable island agency.

References

Carroll, S.R., Garba, I., Figueroa-Rodríguez, O.L., Holbrook, J., Lovett, R., Materechera, S., Parsons, M., Raseroka, K., Rodriguez-Lonebear, D., Rowe, R., Sara, R., Walker, J.D., Anderson, J. and Hudson, M. (2020) ‘The CARE Principles for Indigenous Data Governance’, Data Science Journal, 19, 43. https://doi.org/10.5334/dsj-2020-043

Cabinet Office of Sri Lanka (2024) ‘Sri Lanka’s National Strategy on Artificial Intelligence’, Cabinet decision, 26 August. https://www.cabinetoffice.gov.lk/cab/index.php?Itemid=49&dID=12770&id=16&lang=en&option=com_content&view=article

Faroese Ministry of Children, Education, Research and Culture (2026) UNESCO-tilmæli um etiskar leiðreglur fyri vitlíki. 5 May. https://www.mmr.fo/fo/kunning/alit/unesco-tilmaeli-um-etiskar-leidreglur-fyri-vitliki

Franken, J., Reinhold, T., Reichert, L. and Reuter, C. (2022) ‘The digital divide in state vulnerability to submarine communications cable failure’, International Journal of Critical Infrastructure Protection, 38, 100522. https://doi.org/10.1016/j.ijcip.2022.100522

Government of Iceland (2021) Policy on Artificial Intelligence. https://www.government.is/topics/information-technology/policies/

Government of Ireland (2024) National AI Strategy Refresh 2024. Dublin: Department of Enterprise, Tourism and Employment. https://enterprise.gov.ie/en/publications/national-ai-strategy-refresh-2024.html

Government of Malta (2019) Malta: The Ultimate AI Launchpad: A Strategy and Vision for Artificial Intelligence in Malta 2030. https://oecd.ai/en/dashboards/policy-initiatives/a-strategy-and-vision-for-artificial-intelligence-in-malta-2030

Hua, D. and Brimble, P. (2025) ‘Bridging the divide: building Pacific agency in the AI era’, Devpolicy Blog, 11 July. Australian National University Development Policy Centre. https://devpolicy.org/bridging-the-divide-building-pacific-agency-in-the-ai-era-20250711/

IEA (2025) Energy and AI. Paris: International Energy Agency. https://www.iea.org/reports/energy-and-ai

Ismail, Z. (2019) Public Sector Reform and Capacity Building in Small Island Developing States. K4D Helpdesk Report 583. Birmingham: University of Birmingham. https://gsdrc.org/publications/public-sector-reform-and-capacity-building-in-small-island-developing-states/

Isle of Man Government (2025) ‘Digital Isle of Man to lead early work to establish National AI Office’, 16 October. https://www.gov.im/news/2025/oct/16/digital-isle-of-man-to-lead-early-work-to-establish-national-ai-office/

Jamaica National AI Task Force (2025) National Artificial Intelligence Task Force Policy Recommendations. Kingston: Office of the Prime Minister. https://opm.gov.jm/wp-content/uploads/2025/02/National-Artificial-Intelligence-Task-Force-Policy-Recommendations-Final-1.pdf

Jennings, A. (2026a) AI-slands: Towards an Island-Centred Approach to Artificial Intelligence. AI-slands Working Paper No. 1. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-1.html

Jennings, A. (2026b) AI and Island Cultures: Preservation, Extraction and the Right to Refuse. AI-slands Working Paper No. 2. Institute for Northern Studies, University of the Highlands and Islands. https://ai.insularworlds.co.uk/working-paper-2.html

Joshi, P., Santy, S., Budhiraja, A., Bali, K. and Choudhury, M. (2020) ‘The State and Fate of Linguistic Diversity and Inclusion in the NLP World’, in Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pp. 6282–6293. Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-main.560

Ministry of Culture and Business Affairs, Iceland, Almannarómur and Miðeind (2024) The Icelandic Approach: Preserving and Revitalizing Linguistic and Cultural Diversity in AI. Reykjavík: Government Offices of Iceland. https://www.stjornarradid.is/library/?itemid=e30de195-79d3-11ef-b888-005056bcde1f

Ministry of Public Administration and Artificial Intelligence, Trinidad and Tobago (2026) National AI Assessment. https://mpaai.gov.tt/

Ministry of Business, Innovation and Employment (MBIE) (2025) New Zealand's Strategy for Artificial Intelligence: Investing with Confidence. Wellington: New Zealand Government. https://www.mbie.govt.nz/business-and-employment/economic-growth/digital-policy/new-zealands-ai-strategy-investing-with-confidence

Nikulásdóttir, A.B., Guðnason, J., Ingason, A.K., Loftsson, H., Rögnvaldsson, E., Sigurðsson, E.F. and Steingrímsson, S. (2020) ‘Language Technology Programme for Icelandic 2019–2023’, in Proceedings of the Twelfth Language Resources and Evaluation Conference, pp. 3414–3422. Marseille: European Language Resources Association. https://doi.org/10.48550/arXiv.2003.09244

OECD.AI (2026) ‘National Artificial Intelligence Strategy (ENIA)’, OECD.AI Policy Navigator. https://oecd.ai/en/dashboards/policy-initiatives/national-artificial-intelligence-strategy-enia

Republic of Cyprus (2020) National Strategy for Artificial Intelligence: Actions for the Utilisation and Development of AI in Cyprus. https://www.gov.cy/dmrid/en/documents/national-strategy-for-artificial-intelligence/

Smart Nation Singapore (2023) National AI Strategy 2.0: AI for the Public Good, for Singapore and the World. Singapore: Government of Singapore. https://go.gov.sg/nais2023

Smart Nation Singapore (2026) Update to National AI Strategy. Singapore: Government of Singapore. https://www.smartnation.gov.sg/initiatives/national-ai-strategy/

Spano, I. and Zhang, Y. (2025) ‘Indigenous data sovereignty in intangible cultural heritage governance: A complementary approach to public–private partnerships’, International Journal of Cultural Property, 32(2), pp. 167–193. https://doi.org/10.1017/S0940739125100064

State of Hawaiʻi Office of Enterprise Technology Services (n.d.) Data and AI Guidelines. https://data.hawaii.gov/policies-and-documents/

UNCTAD (2022) Digital Economy in Small Island Developing States. TrainForTrade. https://tft.unctad.org/en/2022/11/12/digital-economy-in-small-island-developing-states/

UNDP (2026) ‘Mauritius launches National AI Strategy and FAIR Guidelines supported by UNDP’, 10 April. https://www.undp.org/mauritius-seychelles/press-releases/mauritius-launches-national-ai-strategy-and-fair-guidelines-supported-undp

UNDP, CAF, IDB and Government of Trinidad and Tobago (2024) SIDS 2.0: Defining the Pathway for Small Islands DIGITAL States in the Caribbean: Summary Outcome Statement. High Level Conference, Port of Spain, 30 April–1 May. https://www.undp.org/sites/g/files/zskgke326/files/2024-06/sids-2.0-initiative-digital-pathways-for-the-caribbean-final-outcome-statement-v02.pdf

UNESCO (2021) Recommendation on the Ethics of Artificial Intelligence. Paris: UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137

UNESCO (2025) ‘Maldives’, Global AI Ethics and Governance Observatory. https://www.unesco.org/ethics-ai/en/maldives

UNESCO and Innovation for Policy Foundation (2022) Multistakeholder AI Development: 10 Building Blocks for Inclusive Policy Design. Paris and Ebene: UNESCO and i4Policy. https://unesdoc.unesco.org/ark:/48223/pf0000382570