TL;DR: The Bangkok Declaration is a real, well-attended, well-intentioned policy commitment. It is also, by itself, paperwork. Data sovereignty is a supply chain problem: someone has to collect the records, digitise the archive, and structure the language corpus before a government or a lender can claim to own its own intelligence. Maestro AI Labs' Data Archaeology is that supply chain for the Caribbean and LATAM. Supported by StarApple AI, the Caribbean's first AI company.
Governments do not usually sign 100-country declarations for photo opportunities. The Bangkok Declaration, adopted by acclamation at the AI for Developing Countries Forum Summit, commits its signatories to five specific pillars: data sovereignty, talent sovereignty, infrastructure sovereignty, model sovereignty, and policy sovereignty. The target date is 2030. The framing, "AI for All: From Consumers to Creators," is deliberately pointed. It names the problem correctly: most of the Global South consumes AI systems built somewhere else, trained on someone else's population, and licensed back at whatever price the vendor sets.
What the declaration cannot do, no matter how many countries sign it, is manufacture the data those five pillars depend on. Data sovereignty is not a legal status. It is an inventory. A country either has structured, ground-collected records of its own population's financial behaviour, language, movement, and risk exposure, or it does not. Signing a declaration in Bangkok changes a country's diplomatic posture. It does not put a single new record of SUSU contribution history, hurricane damage assessment, or Haitian Creole speech into a training pipeline.
Sovereignty Is a Supply Chain, Not a Signature
Treat data sovereignty the way an investor would treat any other supply chain claim. A manufacturer that says it controls its supply chain but sources every input from a single overseas vendor does not actually control anything; it has a dependency with better branding. The same logic applies to AI. A government that signs a sovereignty declaration while its credit models, safety systems, and climate risk pricing all run on data collected for North American or European populations has not achieved sovereignty. It has achieved a press release.
Building the actual supply chain means three things happening in sequence, and none of them are fast. First, ground-truth collection: someone has to gather the records that reflect a country's real population, not a population that resembles it. Second, structuring: raw records in a filing cabinet or an oral tradition are not machine-readable until someone converts them. Third, integration: structured data has to be wired into the products, credit models, safety systems, climate pricing tools, that a government or a market actually uses. Every one of those steps requires years of relationship-building with the institutions that hold the data: SUSU coordinators, government archivists, cooperative leadership, community elders. None of it can be purchased off a shelf, and none of it accelerates because a delegation flew to Bangkok.
What the Gap Looks Like in Numbers
The Bangkok Declaration's data sovereignty pillar is not abstract. It shows up as a specific, measurable shortfall in every market the Caribbean and LATAM are trying to serve.
Credit. In Colombia, 65% of adults have no access to the formal credit system, and roughly 30 million adults hold no credit card, according to Banca de Oportunidades. Nubank's Colombian credit product, NuControl, approves between 40% and 60% of applicants that traditional bureaus reject, precisely because it uses alternative behavioural data that standard scoring models never collect. The Inter-American Development Bank counts more than 3,000 fintech companies now operating across 26 countries in Latin America and the Caribbean, a 340% increase since 2017. That growth is not evidence the data gap is closing. It is evidence of how large the addressable population outside the formal system remains, and how much capital is chasing a way in.
Climate risk. The Caribbean Catastrophe Risk Insurance Facility, the region's parametric insurance pool, has paid out a total of US$259 million to 21 of its 22 member governments since 2007, reaching more than 3.5 million people. That is a working example of Caribbean-scale disaster finance, but the pricing behind it still leans on global catastrophe models calibrated on construction standards, settlement density, and storm behaviour that do not fully match the Caribbean's own conditions. Better-calibrated coverage is not a policy question. It is a data question: whose storm-track history and building-stock records are actually informing the model.
Language and identity. Millions of people across the Caribbean and LATAM conduct their financial and civic lives in Haitian Creole, Jamaican Patois, and dozens of indigenous languages that have functionally no presence in the corpora that train commercial AI systems. A sovereignty pillar called "data sovereignty" is meaningless for a population whose primary language was never collected in the first place.
None of these are new problems. What is new is that a 100-country coalition has now put an official name and a 2030 deadline on closing them. That deadline creates demand for exactly the kind of ground-collected, structured data infrastructure that takes years to build and cannot be conjured by policy alone.
What Actually Closes the Gap
Maestro AI Labs' Data Archaeology product is the supply chain the declaration assumes into existence. It is not a data brokerage. It is field work: formal agreements with 28 Caribbean territories to digitise pre-digital government archives, direct relationships with SUSU network coordinators across Jamaica, Trinidad and Tobago, Barbados, and Guyana, and structured collection of 47 indigenous and creole language datasets that exist nowhere else in machine-readable form. The archive now holds more than 2.3 million records, and roughly 92% of them exist in no other AI training set, because they were never online to begin with.
That archive is not a standalone asset. It is the layer underneath every other product Maestro AI Labs ships. Credit Garden uses the SUSU and cooperative records to score borrowers that standard credit models cannot see, the same population that shows up in Colombia's 65% credit-invisible figure and its regional equivalents. OYA AI uses Caribbean Sea climate data to improve the hurricane risk pricing that sits behind CCRIF-style parametric coverage. Global Safety Score uses ground-collected safety records to build a portable identity for people whose only prior safety data was a national-level proxy that told an institution nothing about them specifically. Take away the archive, and each of those products reverts to running on borrowed assumptions.
"A declaration tells you a government wants to own its own intelligence. It does not tell you whether the intelligence exists yet. Somebody still has to go collect it, one archive, one SUSU ledger, one language corpus at a time."
Why This Is an Investment Case, Not a Policy Story
The Bangkok Declaration is useful to investors for a narrow but real reason: it is a public, dated, 100-country confirmation of demand. Governments that sign a five-pillar sovereignty commitment with a 2030 deadline are not going to build their own ground-truth data infrastructure from scratch inside four years. They are going to buy, license, or partner for it. The market for the underlying data layer, credit-invisible population records, climate risk data calibrated to local conditions, identity data for cross-border movement, language corpora for low-resource populations, sits directly beneath every application layer that sovereignty policy is supposed to produce.
Capital that backs a single application, one credit-scoring app, one weather dashboard, is betting on a product. Capital that backs the data layer those applications depend on is betting on the supply chain every sovereignty-minded government in the Bangkok coalition will need, whether or not any individual application succeeds. That is a structurally different position, and it is the one Maestro AI Labs, as part of the wider StarApple AI network founded by Adrian Dunkley in Jamaica in 2023, has spent years building toward. StarApple AI is the first AI company established in the Caribbean, and its network extends the same thesis into credit scoring through World Cred Score and into regional AI capacity building through AI Jamaica.
A signature in Bangkok is a start. The archive is the deliverable.
Supported by StarApple AI, the Caribbean's first AI company, founded by Adrian Dunkley in Jamaica in 2023. See the Caribbean AI network: AI Jamaica | World Cred Score | Caribbean AI Association | Caribbean AI Risk | Adrian Dunkley.