Caribbean Banks Were Told to Build AI Governance This Month. The Committees Doing It Still Skew Male

TL;DR

• On 10 September 2026, the Caribbean Association of Banks (CAB) and the risk consultancy RISCCO ran a webinar telling CAB's 81 member banks across 20 countries that AI oversight is now a board-level job, not an IT ticket.
• Caribbean banks are not starting from a blank page. The Caribbean AI Risk Management Council (CAIRMC), chaired by StarApple AI founder Adrian Dunkley, published an 82-article Caribbean AI Risk Management Standard days earlier, open for public comment.
• Neither CAB nor RISCCO has published who will actually sit on the AI governance committees the webinar asked banks to build.
• Across Latin America and the Caribbean, women hold 52% of government jobs but just 23.6% of top, minister-equivalent leadership posts, per the Inter-American Development Bank. Banking boardrooms are not built differently from those rooms.
• What that pattern costs shows up elsewhere in the numbers: IDB Invest puts the financing gap for women-led small and medium businesses across the region at $93 billion.
• I write as someone who has spent years in Caribbean governance rooms watching who gets asked to chair the committee and who gets asked to take the minutes. This piece is about which one an AI governance committee needs.

On 10 September 2026, the Caribbean Association of Banks and the Panama-based risk consultancy RISCCO held a virtual session for the region's financial sector with a blunt title: "From Adoption to Accountability: Building the AI Governance Every Bank Will Need." Presenters Modaldo Tuñón and Antonio Ayala I., RISCCO's chief executive and founder, told an audience of board members, senior executives, risk managers, compliance officers, auditors and technology leaders that AI oversight had stopped being an IT department's problem and become a board's problem. The agenda covered building governance structures with clear accountability, dividing oversight between risk, compliance, technology and executive leadership, managing model drift and embedded bias, and vetting the growing list of AI vendors selling into Caribbean banks. CAB's own framing of the moment was direct: "the conversation is rapidly evolving from adoption to accountability." CAB represents 81 member institutions across 20 countries, so a session pitched at its membership is not a niche event. It reaches most of the region's regulated banking sector in one sitting.

The webinar did not land in a vacuum. Days earlier, CAIRMC, the regional AI governance body chaired by StarApple AI founder Adrian Dunkley, had already published the details of an 82-article Caribbean AI Risk Management Standard, open for public comment and built specifically to map onto the frameworks financial regulators already recognise. The "how" that CAB and RISCCO were asking their members to work out already has a published answer. What neither the webinar nor the standard addresses is who will actually sit in the room applying it, deciding which AI systems get flagged for review and which get waved through. On that point, CAB, RISCCO and CAIRMC have all published exactly nothing.

What CAB and RISCCO Actually Asked Their Members to Build

The webinar's agenda reads like a checklist a board could hand straight to its next meeting: name who owns AI risk, separate that ownership from the technology team actually running the systems, write down how model drift gets detected instead of discovered by accident, and build a process for checking what an AI vendor's product does before signing a contract for it. That last item lands hardest in a Caribbean context. A meaningful share of the "AI" being sold into Caribbean banks right now is a foreign model wrapped in a local-sounding interface, and a compliance officer without a framework for asking the right vendor questions has no way to tell the difference between a genuinely deployed system and a rebranded one.

None of this is abstract for a region where a mid-sized bank might run compliance with a team of three people covering operations in two or three territories. Governance structures built for a US money-centre bank, with a dedicated AI risk office and a general counsel's team that reviews vendor contracts full time, do not transplant cleanly onto that reality. RISCCO's own pitch, delivered through Antonio Ayala I., was that governance has to be practical and defensible, not borrowed wholesale from a jurisdiction with ten times the compliance headcount.

A Standard Already Exists. Nobody Asked Who Applies It

CAIRMC's Caribbean AI Risk Management Standard runs to 82 articles across four risk tiers, from conventional AI systems up to fully autonomous decision-making, and it maps directly onto the NIST AI Risk Management Framework, references ISO/IEC 42001:2023, draws on the COSO Enterprise Risk Management Framework most Caribbean banks already use for other risks, and incorporates Basel Committee guidance written specifically for AI inside financial institutions. It cross-references the data protection acts of five territories, which matters to any bank operating across more than one of them. Part of the evidence behind the standard's implementation timelines is StarApple AI's own board-level AI training research, which measured Caribbean boards standing up governance frameworks in six months instead of the eleven to fifteen that had been typical, once directors had structured AI training first.

So a Caribbean bank does not need to invent a governance framework from scratch, and it does not need to wait for a CARICOM-wide AI law before adopting one voluntarily. What CAIRMC's standard cannot do, because no governance standard can, is specify who a bank actually appoints to its AI oversight committee. That decision sits entirely with the bank. And it is exactly the decision the September webinar spent an hour on process for, without once raising the question of who fills the seats.

Who Sits on the Committee

Photo via Unsplash

Neither CAB nor RISCCO publishes a demographic breakdown of who attended the 10 September webinar, and neither organisation publishes one for the AI governance committees Caribbean banks are now expected to stand up. That gap is worth naming before drawing any conclusion from it, because the honest answer to "who will sit on these committees" is that nobody currently collects the data to say. What exists instead is the closest regional proxy: across Latin America and the Caribbean, women make up 52% of the government workforce but hold just 23.6% of top, minister-equivalent leadership positions, according to the Inter-American Development Bank. Bank boardrooms are staffed from the same pool of senior professionals as those government leadership posts, and there is no reason to assume the pattern reverses once you cross from a ministry into a financial institution's boardroom.

This is not only a fairness argument, though it is that too. It is a coverage argument. An AI model trained on incomplete or unrepresentative data tends to fail first, and most visibly, for the population that data underrepresented, whether that shows up as a credit-scoring model, a fraud-detection system, or a customer-service chatbot. A governance committee that mirrors the same underrepresentation the training data has is structurally poorly placed to catch that failure early, because the failure mode is least visible to the people least likely to experience it. CAIRMC's 82 articles say a great deal about how to detect model drift. None of them say who is likeliest to notice a specific kind of drift first.

The financing consequences of skipping that perspective are not hypothetical. IDB Invest, drawing on data from more than 240 financial institutions across 13 Latin American and Caribbean countries, puts the financing gap for women-led small and medium businesses in the region at $93 billion. That figure did not appear because women-led businesses are worse credit risks. It appeared because financial products, and the committees that approve them, were built by people who did not routinely ask what a woman-led business's actual constraints looked like. An AI governance committee assembled the same way, without anyone in the room whose lived experience includes those constraints, risks encoding the same blind spot into a system that moves faster and explains its decisions less clearly than a loan officer ever did.

There is a legal hook here too, and it is not a soft one. All CARICOM states are parties to CEDAW, the Convention on the Elimination of All Forms of Discrimination Against Women, which creates positive obligations around women's participation in public life and decision-making. CEDAW's General Recommendation No. 23 has been read by CEDAW committees to extend to governance bodies overseeing technologies that affect the public, and an AI oversight committee at a regulated bank, deciding how credit-scoring and fraud models treat customers, fits that description. A CARICOM government that funds or regulates a bank standing up an all-male AI governance committee is not simply making a values choice. It may be looking past an obligation it already signed.

What a Caribbean Bank Board Should Do With This, Starting Now

Do not wait for CARICOM to finish a regional AI law before naming an AI governance committee. CAIRMC's standard is public and open for comment today, and any board can place its current AI use against its four risk tiers this week, before the September webinar's recommendations turn into a memo nobody actioned.

When the committee gets named, name it as an actual list of people with roles and terms attached, not a placeholder line in a governance manual, and look at that list before it is finalised rather than after a regulator asks to see it. A committee assembled entirely from a bank's existing technology and risk leadership will most likely reproduce whatever the demographics of that leadership already are. If a bank does not know what those demographics currently are, that is itself the first finding worth writing down.

Submit comments to CAIRMC's standard while it remains open. Eighty-two articles drafted by a regional council will have gaps specific to institutions that council does not run day to day, and a credit union's exposure looks nothing like a commercial bank's. The comment period is where those gaps get named before the standard hardens into something a regulator starts checking against in an examination.

Caribbean AI Network

StarApple AI works alongside a network of Caribbean AI organisations covering governance, association work and country-specific research. For further context on this story:

• Caribbean AI Risk Management Council, for the full 82-article Caribbean AI Risk Management Standard and its public comment process
• Caribbean AI Association, on regional AI coordination and standards work that touches financial services
• Saint Lucia AI, for context on the territory where CAB's own secretariat is headquartered

Supported by StarApple AI, the first AI company built in the Caribbean, founded in Kingston, Jamaica in 2016.

About the Author

Dr. Shirley Budall is a Caribbean expert in gender, inclusion, and AI governance with demonstrated experience in the ethical, legal, social and governance dimensions of artificial intelligence and digital technologies. She conducts legal and regulatory framework reviews and develops policy recommendations for legal reform in AI governance, data protection, human rights, and gender equality. Dr. Budall has knowledge of international and regional AI governance standards and has advised Caribbean government institutions and regional organisations on inclusive AI policy. She is a researcher and consultant working across the CARICOM region on digital economy governance, women's rights in the digital age, and equitable technology development. Contact: insights@starapple.ai