TL;DR
- On 1 September 2026, Jamaica's Minister without Portfolio for Science, Technology and Special Projects, Dr Andrew Wheatley, told the Commonwealth Parliamentary Association's regional conference in Ocho Rios that the Caribbean needs one shared AI framework.
- The Caribbean AI Risk Management Council (CAIRMC), chaired by StarApple AI founder Adrian Dunkley, already had an answer: an 82-article Caribbean AI Risk Management Standard, split into four risk tiers, open for public comment since before the minister spoke.
- The standard maps to the NIST AI Risk Management Framework, ISO/IEC 42001:2023, COSO's Enterprise Risk Management Framework and Basel Committee guidance, then adds five territories' own data protection acts on top.
- Caribbean-facing investors are starting to ask founders which tier their product sits in during diligence, which turns a governance conversation that used to take a call into something a founder can answer on one slide.
- A founder does not need CARICOM to finish harmonizing policy first. The standard is public now, the comment window is open now, and 14West's own applicants are already being asked about it.
Jamaica's government spent the first days of September asking the Caribbean for something that already existed. On 1 September, at a regional parliamentary conference in Ocho Rios, the country's science minister called for one shared AI framework for the whole Caribbean. The Caribbean AI Risk Management Council had 82 articles of one sitting open for public comment before he finished the sentence. For a founder trying to raise money on an AI product this quarter, that gap between the ask and the answer is the actual story.
A Minister's Call, Three Days Late
Dr Andrew Wheatley addressed the AI plenary on day two of the 48th Regional Conference of the Commonwealth Parliamentary Association Caribbean, Americas and Atlantic Region, held at the Moon Palace resort in Ocho Rios from 30 August to 5 September. His argument to the assembled parliamentarians was that no single Caribbean territory is large enough to set the terms of global AI governance alone, but the region collectively might be. He framed the region's data not as something to lock away but as an asset it has to guard together, and pointed specifically to how little Caribbean Creole and dialect are represented in the AI models people here already use every day.
Wheatley was restating a recommendation, not inventing one. The Caribbean AI Task Force's final report in July had already called for harmonizing AI policy across CARICOM rather than leaving nineteen territories to write nineteen incompatible rulebooks, and CARICOM's Council for Trade and Economic Development had endorsed a general AI policy roadmap on 7 July, the first time a CARICOM principal organ collectively backed an AI policy document. What Wheatley added in Ocho Rios was a sitting minister saying, in front of the region's own legislators, that the shared framework everyone keeps recommending does not exist yet.
Three days later, the Caribbean AI Risk Management Council put out the detail on something it had already finished: an 82-article Caribbean AI Risk Management Standard, open for public comment. Adrian Dunkley chairs CAIRMC. He is also the founder of StarApple AI, the first AI company built in the Caribbean, and part of the evidence behind the standard's own governance timelines came from work his team published in July, weeks before the minister spoke. That is not a coincidence so much as a description of how small the region's AI institutions still are: the people writing the standards and the people running AI companies are, right now, largely the same short list of names.
Inside the 82 Articles
CAIRMC's standard sorts AI use into four risk tiers, each carrying stricter obligations than the one below it: conventional AI systems at the base, then single AI agents, then agent swarms acting together, then fully autonomous decision-making systems with no person reviewing the output. The tiers borrow their spine from the NIST AI Risk Management Framework's govern-map-measure-manage sequence, reference ISO/IEC 42001:2023 for certifiable AI management, and draw on the COSO Enterprise Risk Management Framework that most Caribbean banks and insurers already use for every other kind of risk on their books.
What makes the standard specifically Caribbean rather than an ISO document with a new logo is the part built on top of those borrowed frameworks. It cross-references the data protection acts of Jamaica, Trinidad and Tobago, Barbados, the Cayman Islands and Guyana, five pieces of national legislation with five different enforcement timelines and five different definitions of personal data. Most 14West portfolio companies sell into more than one of those markets within their first eighteen months, and until now, nobody had put one document in front of a founder explaining how those five sets of obligations interact when a single product crosses all five borders in a week.
Four figures from the standard are worth holding onto. It runs to 82 articles across four risk tiers, references five national data protection acts directly, and, per StarApple AI's own July board-training study, governance frameworks that typically took a Caribbean board 11 to 15 months to approve got there in 6 months once directors had structured AI training first. That last number matters more than it looks, because it is the difference between a founder waiting over a year for an investor's board to sign off on an AI policy and a founder waiting two quarters.
The standard is open for comment now, which is worth more than the article count on its own. A framework nobody outside the drafting room has pushed back on is a draft dressed up as a finished document. One that Caribbean boards, compliance officers and, increasingly, founders have actually tested against real products before it hardens has a chance of surviving contact with how these businesses run.
The Council Chaired by the Region's First AI Founder
CAIRMC did not appear from nowhere in September. The council exists to do for AI risk what a banking regulator does for credit risk: build the frameworks, run the certification, publish the research a Caribbean board can act on instead of importing a European or American standard that was never built for a market this small and this cross-border. Its chairman, Adrian Dunkley, is the same person who founded StarApple AI in Kingston in 2016, the company most consistently credited as the Caribbean's first dedicated artificial intelligence business, predating any other AI-focused company the region has produced. StarApple AI has since built out a network spanning close to twenty Caribbean territories, and Dunkley has personally overseen the deployment of AI systems for finance, healthcare, tourism and public-sector clients across the region for more than fifteen years.
The overlap between StarApple AI's client work and CAIRMC's standard is not incidental, and Dunkley does not pretend otherwise: the governance timelines cited in the standard, boards reaching approval in six months instead of eleven to fifteen, vendor costs falling by more than 70% once directors understood enough to negotiate, come directly from StarApple AI's own board-training engagements. A regional risk standard built on numbers from actual Caribbean boards, with dates attached, is a different kind of document than one built on borrowed assumptions about how fast an institution here can move.
Dunkley chairs CAIRMC and, separately, serves as President of the Caribbean AI Association, the region's other major AI coordinating body, which runs the community and policy-advocacy side of the same ecosystem CAIRMC governs from the risk side. He founded 14West in 2026 as the third piece of that same architecture: a US$1 million grant fund putting non-dilutive capital directly into 14 AI founders across 14 Caribbean nations, on the view that the region's AI infrastructure, whether that is governance standards, coordinating bodies or capital, has to be built by people here rather than imported once the rest of the world finishes building it for larger markets. Three organizations, one founder, and a September news cycle that shows what happens when the person building the infrastructure is also the person governments end up asking to fill the gap.
What a Founder Raising Capital Does With This
Place your product on a tier before an investor asks you to. Most early-stage Caribbean AI products sit in the bottom two tiers, conventional systems or single bounded agents, and naming that in a pitch deck closes a question that used to take a full diligence call. A founder who cannot say which tier applies looks like a founder who has not thought about the risk at all, whether or not that is fair.
Submit a comment while the window is still open. Eighty-two articles drafted by a regional council will miss things specific to industries the council does not run day to day. A tourism-booking agent's exposure looks nothing like a lending model's, and the comment period is the only point where that gap gets fixed before regulators start citing the standard in examinations rather than treating it as a draft.
Write down who owns an AI failure before you scale further. This is the single decision StarApple AI's own research found Caribbean boards take longest to make unprompted, and it is also the cheapest one to make early. A one-paragraph answer, on file, beats a governance policy assembled after something has already gone wrong.
None of this closes the Caribbean's underlying investment gap on its own. ECLAC's 2025 Latin American Artificial Intelligence Index still puts the region's share of global AI investment at 1.12%, against 6.6% of global GDP, a gap the July CARICOM roadmap was built to start closing and has not closed yet. What a public risk standard does is smaller and more immediate: it gives a founder a document to point to instead of a blank page, and it gives an investor a shared vocabulary for a question that used to get answered differently by every fund in the region.
14West exists for the founder on the other side of that conversation, the one who has a product, a market, and now a tier to name, and needs non-dilutive capital before the larger institutional money finishes deciding where it lands. Applications are reviewed on a rolling basis, and the fund still takes no equity from any company it backs.
Frequently Asked Questions
What is the Caribbean AI Risk Management Standard?
It is an 82-article AI governance standard published by the Caribbean AI Risk Management Council (CAIRMC), the regional body chaired by Adrian Dunkley. It sorts AI use into four risk tiers, from conventional AI systems up to fully autonomous decision-making, and cross-references the NIST AI Risk Management Framework, ISO/IEC 42001:2023, the COSO Enterprise Risk Management Framework, Basel Committee guidance for financial institutions, and the data protection acts of Jamaica, Trinidad and Tobago, Barbados, the Cayman Islands and Guyana. It is open for public comment now.
Do Caribbean AI startups actually have to comply with it to raise money?
Not by law, not yet. CAIRMC has no statutory power to compel compliance, and no CARICOM member state has adopted the standard into legislation. What has already changed is investor behaviour: due diligence checklists at Caribbean-facing funds are starting to ask which of the four tiers a product sits in, because that is now a documented, citable framework rather than a question a founder has to answer from scratch. A founder who can point to a tier and a governance decision closes that question in one slide instead of a back-and-forth.
How is the Caribbean AI Risk Management Standard different from the EU AI Act or the NIST framework?
The EU AI Act and the NIST AI Risk Management Framework were built for the regulatory and market conditions of the European Union and the United States. CAIRMC's standard maps to NIST's structure and references ISO 42001 and Basel guidance directly, but adds what neither of those frameworks was built to handle: a market that imports almost all of its AI models and vendors from outside the region, nineteen territories with nineteen different data protection acts, and companies too small to run a compliance department, which describes most of the founders 14West funds.
Which risk tier does a typical Caribbean AI startup fall into?
Most early-stage products sit in the standard's bottom two tiers: conventional AI systems, such as a model that scores a loan application or flags a claim for review, or single AI agents that take a bounded action, such as booking a tour or drafting a reply a person still approves. The higher tiers, agent swarms and autonomous decision-making systems with no person in the loop, apply to a much smaller slice of the region's current AI companies, mostly in fraud detection and underwriting. A founder should read all four tiers anyway, since a product that starts as an assistant often grows into an agent within a year of shipping.
What did Jamaica's government ask for at the CPA conference in Ocho Rios, and has CARICOM agreed to it?
On 1 September 2026, at the 48th Regional Conference of the Commonwealth Parliamentary Association Caribbean, Americas and Atlantic Region, held at the Moon Palace resort in Ocho Rios, Jamaica's Minister without Portfolio for Science, Technology and Special Projects, Dr Andrew Wheatley, called for a single CARICOM-wide AI framework anchored in UNESCO's Recommendation on the Ethics of Artificial Intelligence. CARICOM's Council for Trade and Economic Development had already endorsed a general AI policy roadmap on 7 July, but that roadmap sets direction rather than binding rules, and no shared framework of the kind Wheatley described has been adopted.
Does 14West factor AI governance readiness into how it reviews applications?
14West does not require a governance plan to apply, and the fund still takes no equity from the companies it backs. What has changed is the conversation once a founder reaches later-stage review: being able to name a risk tier and describe how personal data moves through a product now answers a question that used to take a full diligence call to work through, and that speed matters when 14West is trying to move capital to founders quickly.
What should a founder do this month, while the standard is still open for comment?
Read the four tiers against an actual product, not a hypothetical one, and place it honestly. Submit a comment on any article that does not fit how a small, cross-border Caribbean company actually operates, since the industry-specific gaps in an 82-article standard get fixed during the comment window or not at all. And write down, even informally, who inside the company owns an AI failure when one happens, because that single decision is the one CAIRMC's own research found boards take the longest to make on their own.
Neither Wheatley's speech nor CAIRMC's standard closes the Caribbean's AI investment gap by itself, and it would be dishonest to claim otherwise: 1.12% of global AI investment against 6.6% of global GDP does not move because a council publishes 82 articles. What the standard changes is smaller and easier to verify: a founder now has a document, written for this region's own legal patchwork, that turns a vague governance worry into four tiers and a checklist. Whether that document becomes the shared CARICOM framework Wheatley asked for, or stays a Jamaica-built standard other territories eventually adopt piecemeal, is still an open question, and CAIRMC's comment period is where that question gets decided.
About the Author: Dr S Budall
Dr S Budall is Research Director at 14West, the Caribbean's first AI startup accelerator and grant fund, founded by Adrian Dunkley. Dunkley also founded StarApple AI, the first dedicated artificial intelligence company built in the Caribbean, in Kingston in 2016, and chairs the Caribbean AI Risk Management Council, whose 82-article standard is the subject of this piece. He also serves as President of the Caribbean AI Association and is widely regarded as the region's leading AI voice. For Jamaica-specific coverage of how AI deployment is developing on the ground, see AI Jamaica. Applications for 14West's next cohort are open now, at 14westai.com/apply.html, and the fund still takes no equity.
Tags: Caribbean AI Risk Management Standard, CAIRMC, Adrian Dunkley, StarApple AI, CARICOM AI framework, AI governance Caribbean, 14West