The Case for a CARICOM AI Charter

Note from the President

In two months our Charter Working Group will publish the final regional AI Charter for adoption by CAIRA member countries. The draft has been through five rounds of public comment since February. More than four hundred members and twenty seven institutions have contributed. Andre Beaumont, who has led the drafting, has written this month's feature on what the Charter is and why it matters. The piece doubles as the public introduction to the document, and member comment on the last open sections is welcome until July 31.

We are deep into hurricane season. The systems forming in the deep tropics have not, so far this year, intensified the way the early forecasts suggested they might. That is not a guarantee for August. Keep your preparation current.

A short observation. CAIRA's first anniversary is in August. I have written, in the back of a notebook, a few thoughts about what I think we have got right and where I think we have fallen short. I will share those in the next issue, alongside Lisette's industry piece. Until then.

Adrian Dunkley Founder and President, Caribbean AI Association

---

Feature

The Case for a CARICOM AI Charter

By Andre Beaumont

In September of last year, when I joined the CAIRA Charter Working Group, the question put to us was simple to state and difficult to answer. Could the Caribbean, as a region, produce a coherent policy framework on artificial intelligence that the fifteen member states of CARICOM could adopt, that respected the legitimate differences among them, that took seriously the risks of the technology without being captured by them, and that opened more doors than it closed for our economies and our people. The Charter that we will publish in September is our answer. This piece sets out the case for it.

Why a regional charter rather than fifteen national policies.

The default option, and the one several of our governments have begun to pursue, is for each Caribbean state to develop its own national AI strategy and regulatory framework. The instinct is understandable. Sovereignty is real. National contexts differ. The political timeline of policy in each country has its own shape.

There are three reasons we believe a coordinated regional framework is the better path, and the reasons are not the ones often cited.

The first is regulatory burden on the supply side. The Caribbean is a small market in aggregate, and a tiny market when subdivided. A US or European technology provider deciding whether to invest in Caribbean specific compliance work, in product localization, in language support, in customer support staffing for our region, will look at the cost of doing that work fifteen times for fifteen distinct regulatory regimes and conclude, in most cases, that the math does not justify it. The result is not that we get fifteen carefully tailored offerings. The result is that we get one standardized offering, designed for somebody else, with our concerns absent from the design. A common regional baseline is the cheapest way to make ourselves worth designing for.

The second is regulatory burden on the demand side. The Caribbean's small public sectors and small private companies do not have the policy capacity to draft sound AI governance from scratch. They have other work to do. A regional Charter provides a tested, defensible reference point that a Minister of National Security in Grenada or a Permanent Secretary in Antigua can adopt, in whole or in part, without commissioning a six month consultancy. Several of our member states have already told us, on the record, that this is the most useful thing CAIRA could provide them this year.

The third is leverage in external negotiation. When CARICOM speaks with one voice to the European Commission, to the US administration, to the Chinese technology firms, and to the international standards bodies, we are heard. When we speak as fifteen voices on different days with different messages, we are not. The Charter is a vehicle for that single voice on AI matters, in the same way the CARICOM Single Market and Economy has been on trade and the Caribbean Court of Justice has been on jurisprudence.

What the Charter contains.

The document is organized into seven parts. Members can read the draft in full at caribbeanaiassociation.com/charter. The summary is as follows.

Part one. Purposes and principles. The Charter is built around five principles. AI deployment in the Caribbean should serve the public good. It should respect the rights, dignity, and language of Caribbean citizens. It should be transparent in its operation and accountable for its outcomes. It should be subject to human oversight on decisions that affect Caribbean lives. It should support, not displace, the autonomy of Caribbean institutions and Caribbean workers. These principles are not unique to our region. The application of them in our context, the choices about what they mean in practice when made for a population of forty million across multiple small states, is the work of the document.

Part two. Definitions. The Charter defines the terms it uses, in a way that is consistent with the EU AI Act where possible and with our own legal traditions where necessary. The definitions matter, because much of the contention in AI policy globally comes from people using the same words to mean different things. The Charter is explicit.

Part three. Risk categorization. The Charter adopts a risk based approach, with four categories from prohibited to minimal risk. The categorization tracks the EU AI Act in structure to ease compliance for firms operating in both jurisdictions, but the specific allocations differ in three respects. First, the Charter classifies AI systems used in immigration and border control at our region's ports of entry as high-risk, with mandatory human oversight requirements, on the basis that error in this domain has historically had outsized consequences for Caribbean people. Second, the Charter classifies AI systems used in coastal monitoring, disaster prediction, and climate adaptation in a separate "essential public infrastructure" category, with both higher requirements for reliability and explicit support for development. Third, the Charter creates a "cultural and linguistic exposure" classification for AI systems whose primary public use involves Caribbean languages or cultural content, with requirements that include linguistic validation in the relevant Caribbean language varieties.

Part four. Rights and recourse. The Charter establishes a set of rights for Caribbean citizens in their interactions with AI systems deployed by public or large private institutions. The right to know when a decision affecting you was made by an AI. The right to an explanation in plain language. The right to a human review, on request, of any consequential automated decision. The right to have your data processed in accordance with your country's data protection law. The right to a complaint mechanism that is accessible without legal representation. These rights are not novel internationally. The work of the Charter is to make them concrete and enforceable in the Caribbean context.

Part five. Procurement. The Charter provides a model procurement clause that any Caribbean government, parastatal body, or large institution may adopt when buying AI systems. The clause covers transparency, bias testing, language support, data localization where required, and termination rights. The point of the model clause is to remove the procurement re-drafting burden from individual ministries and to create a consistent market signal to suppliers. The Government of Barbados has indicated, on the public record, an intention to adopt the model clause for all AI procurement above a defined threshold within twelve months of the Charter's publication.

Part six. Regional cooperation. The Charter creates a Caribbean AI Coordinating Council, drawn from the relevant ministries of CARICOM member states, with secretariat support from CARICOM and standing technical input from CAIRA. The Council's role is coordinating, not executive. It will publish annual reports on regional AI deployment, convene the regional response to significant cross-border AI incidents, coordinate Caribbean positions at international fora, and maintain the model documents the Charter introduces.

Part seven. Implementation and review. The Charter is designed for staged adoption. A country joining the Charter framework can opt in to the principles immediately, the procurement model within twelve months, the rights and recourse provisions within twenty four, and the high-risk category requirements within thirty six. The full Charter is reviewed every three years.

What the Charter does not do.

It is worth being explicit about the scope, because the Charter is sometimes misunderstood by both supporters and critics.

The Charter is not a binding treaty. It is a framework instrument. Each CARICOM member state retains its sovereign authority over the substance and enforcement of AI policy within its territory. The Charter creates a coordinated baseline that member states may, and we hope will, adopt and elaborate.

The Charter is not a research and development funding mechanism. There are good arguments for a regional AI development fund, but the Charter is not it. That work belongs to the Caribbean Development Bank and to the national development institutions, with CAIRA as a partner rather than a funder.

The Charter is not aimed at the frontier AI labs in the United States and Europe. It is aimed at the deployers and procurers of AI systems within our region, who are the parties whose behaviour our laws can directly affect. We have no illusion that the Charter will, by itself, change the design choices of OpenAI, Anthropic, Google, or Meta. The hope is that the Charter, in concert with EU regulation and other regional frameworks, contributes to a market signal those firms are increasingly responsive to.

Where the open questions remain.

The drafting has been hard. Three areas remain genuinely contested, and member comment in July is most welcome on these.

The first is the treatment of facial recognition in public spaces. The current draft prohibits real-time facial recognition by law enforcement in public spaces, with limited and named exceptions. There is a credible counter-view that the carve-outs should be wider, particularly for serious crime investigation. The argument the Working Group has accepted, provisionally, is that the technology's error rates on Black populations remain too high for routine deployment, and that a tighter standard is consistent with our region's commitments under the Inter-American human rights system. Member comment is invited.

The second is the treatment of AI generated content in elections. The draft requires disclosure of AI generated political content in public campaigns. The point of contention is whether the standard should be disclosure or prohibition for certain categories of synthetic content. Our region's electoral systems are diverse, and the right answer here is not yet settled.

The third is the data flow rules for AI systems serving Caribbean users from foreign jurisdictions. The current draft adopts a light touch, requiring transparency about where data is processed but not requiring local processing. Some members, particularly in health and financial services, have argued for tighter rules. The Working Group is open to revising this section based on member comment in July.

A closing thought.

The Caribbean has long understood that small states are not necessarily weak states, but that the strength of small states lies in coordination, in care with words, and in the patience to do the institutional work that the larger countries are too distracted to do well. The AI Charter is, in that respect, a continuation of work this region has done before in financial services, in disaster management, in maritime law, and in cricket. The instinct is the same. Coordinate where the cost of fragmentation is too high. Speak with one voice where one voice carries. Build the institutions that will be there after the personalities have moved on.

I invite every CAIRA member to read the draft and to send comment. We have built the Charter for you to use. We need your help, in this last open month, to make it worthy of the name.

---

Andre Beaumont is a regulatory affairs consultant based in Bridgetown and chairs the CAIRA Charter Working Group.

Originally published in The Monthly Intelligence Report, July 2025.