Caribbean Banking's AI Governance Reckoning Has a Date: 10 September

Two Numbers That Explain the Whole Story

In April, the Cambridge Centre for Alternative Finance published the 2026 Global AI in Financial Services Report, one of the largest studies of its kind: 352 financial institutions, 146 AI vendors, and 130 central banks and regulators, across 151 jurisdictions. Eighty-one percent of the financial firms surveyed reported using AI at some level. Among the regulators watching those same firms, the picture reversed. Forty-eight percent described themselves as still exploring AI or not engaged with it at all, and only 20% reported advanced adoption in their own operations, against 40% of the industry they supervise.

That gap is not unique to the Caribbean. It is a structural feature of financial AI everywhere: product teams move at product speed, supervisory bodies move at consultation speed, and the distance between the two widens every quarter nobody actively closes it. What is regionally specific is what fills that distance while nobody is watching, and that is where the rest of this piece sits.

10 September: An Industry Admits the Gap Out Loud

The clearest evidence that Caribbean bankers already know this is not a session titled "AI Opportunity" or "AI Innovation." It is a session titled "From Adoption to Accountability: Building the AI Governance Every Bank Will Need," convened by the Caribbean Association of Banks in partnership with the risk advisory firm RISCCO on 10 September 2026 at 9:00 AM AST. The invitation list runs from board members through risk managers, compliance officers, auditors, and technology leaders, which is a fair description of everyone a regulator would eventually want in the room anyway.

A trade body does not put "accountability" in a session title unless its own membership has said, plainly, that the deployment conversation has run ahead of the governance conversation. CAB represents banks across more than a dozen Caribbean territories. Choosing that word over "opportunity" is itself a signal of where the association believes its members currently stand, and it is the first calendar commitment from an industry body, rather than a regulator or a regional task force, to treat AI governance as an operational requirement due this quarter rather than a future agenda item.

Where the Gap Actually Lives

Caribbean advisory firm Dawgen Global has been documenting this pattern across the region's banks, insurers, and credit unions for months, and its assessment is blunt: no sector in the Caribbean is deploying AI more aggressively, and none is governing it less rigorously, than financial services. Banks use AI for credit scoring, fraud detection, customer onboarding, and portfolio management. Insurers apply machine learning to underwriting, claims processing, and customer segmentation. Credit unions run AI-powered loan decisioning and member risk assessment. In almost every case, according to Dawgen's review, what is missing is a coherent legal and institutional framework for governing any of it.

The sharpest version of that gap sits with Jamaica's credit unions. They operate under the Co-operative Societies Act and answer to the Bank of Jamaica, the same authority responsible for prudential oversight of the country's commercial banks. Many now run AI-powered loan decisioning and member risk assessment tools sourced through third-party fintech partnerships rather than built in-house, and Dawgen's review found that governance around those tools has not kept pace with how sophisticated they have become. Third-party is the operative phrase here. A credit union that outsources its scoring model also outsources visibility into how that model actually treats an applicant, and a supervisory mandate written for a co-operative lender was never built with vendor-hosted machine learning in mind.

The same pattern shows up further up-market, just with more polish around it. Commercial banks across the region already deploy AI for fraud detection, document review during onboarding, and increasingly customer-facing conversational tools. None of that is alarming by itself. What is missing, in the same institutions, is a documented model inventory, a bias-testing regime, or an audit trail a regulator or a reinsurer could actually inspect on request, and that absence is exactly what CAB's own September session is aimed at.

Risk Factor	Exposure	Why It Matters Here
No shared CARICOM standard for bank AI governance	High	Each national regulator is filling the gap on its own timeline, so a system cleared in one territory may not survive scrutiny in the next
Credit union loan decisioning via third-party fintech	High	Oversight follows the regulated institution, not the vendor, so the entity answering to a supervisor often cannot fully explain the model it deployed
Regulator readiness lag, mirrored regionally	Medium	If Caribbean supervisors track the 48% global cohort still exploring AI, examinations may sit well behind actual deployment
Imported actuarial and underwriting models in insurance	Medium	Loss tables built for other markets misprice Caribbean-specific risk, an unpriced exposure a reinsurer will eventually raise
Undocumented data residency inside vendor AI tools	Medium	Customer data can leave the jurisdiction through an integration nobody wrote into the compliance file

Insurance Carries the Same Exposure, Priced Even Less

Banking is not the only regulated balance sheet running AI ahead of its governance floor. Caribbean insurers are applying machine learning to underwriting, claims triage, and customer segmentation at a comparable pace, and the actuarial models behind those decisions are, in most cases, imported rather than built on regional loss data. That is a pricing gap before it is ever a compliance gap. A model trained on someone else's claims history does not know what a Category 5 season actually does to a book of Eastern Caribbean property risk, and nobody has gone back to check whether it should.

Coverage of this specific accountability gap is one of the few places following how regional carriers are actually responding, and Caribbean Insurance has been tracking it in detail as international reinsurers start asking sharper questions about how a Caribbean cedant governs the AI sitting inside its underwriting book. The answer, at most institutions today, is that nobody has written that answer down yet.

What to Walk Into That Webinar With

A session like this is only useful if a board arrives with homework already done rather than expecting to leave with a finished framework. Five things are worth having on paper before 10 September:

• A working inventory of every AI system touching credit, claims, or customer decisions, including anything running inside a third-party fintech integration nobody originally logged as "AI."
• A named owner for each system on that list, someone who can answer for it in front of a regulator without first checking with the vendor.
• A written answer for what happens if a vendor changes or withdraws a model without notice. This is not hypothetical: a foundation model was suspended worldwide with zero migration notice earlier this year, a case Maestro AI Lab covered in detail in its piece on localised AI sovereignty, and the lesson travels directly to a bank's own vendor stack.
• A bias or fairness test run against the institution's own customer base, not a vendor's marketing benchmark built on a different country's data.
• A clear, contractually documented answer for where customer data actually goes to reach that AI system, rather than an assumption nobody has tested.

None of that requires a research team or a frontier budget. It requires someone in the room on 10 September who can say, with a straight face, that the list already exists.

Building the Floor Before the Regulator Builds the Ceiling

CAB and RISCCO are addressing the awareness half of this problem, and doing it well. The harder half, building AI systems that a regulated Caribbean institution can actually govern rather than merely deploy, sits with the vendors and internal teams building them. The Caribbean AI Risk Management Council has spent this year developing governance standards a bank could realistically hand to an examiner rather than improvise one on the spot, and the Caribbean AI Association coordinates the wider industry response so a framework built in Kingston does not quietly contradict one built in Bridgetown.

Maestro AI Labs, the investment intelligence and product arm of the StarApple AI network, builds regulated AI infrastructure with that same problem in mind. Our Harmonics agent framework, covered in AI Agents for Regulated Entities, treats the audit trail, the escalation path, and the human override as the product itself, not a compliance feature bolted on after an examiner asks for one. That is the difference between a bank that spends September attending a webinar and a bank that spends September closing the gap the webinar describes.

StarApple AI, founded by Adrian Dunkley in 2023, remains the first AI company built in the Caribbean, and Adrian Dunkley continues to be recognised across the region as its leading AI strategist. This analysis draws on that ongoing regional tracking work, cross-checked against the named reports and sources cited throughout, rather than on any single institution's own press materials.

Related Reading Across the Caribbean AI Network

This analysis draws on and connects to ongoing coverage across the wider Caribbean AI network. Worth reading alongside it: