The Caribbean Wrote Its AI Rulebook in a Year. Nobody Has Said Who Enforces It

Two governance moments landed in the same month, and almost nobody outside the rooms where they happened noticed how close together they fell. On July 7, 2026, CARICOM's Council for Trade and Economic Development on Information and Communication Technologies, known as COTED-ICT, formally endorsed the UNESCO Caribbean AI Policy Roadmap at its 126th special meeting. Sixteen days later, on July 23 and 24, the Caribbean Telecommunications Union's Caribbean AI Task Force presented its final report at the first Caribbean AI Forum, hosted at the University of the West Indies St Augustine campus under the theme "AI for Caribbean Transformation: Governance, Innovation and Resilience for a Shared Digital Future."

I welcomed both. I have said in boardrooms across the region, and I will say it here, that a Caribbean that lets AI rules get written entirely in Washington, Brussels, and Beijing is a Caribbean that will spend the next decade complying with standards it had no hand in shaping. Two serious efforts to write our own position, in the same month, is not a coincidence. It is momentum. What it is not, yet, is enforcement.

A Year of Real Work, Compressed Into Two Documents

The Task Force's mandate traces back to October 2025, when the CTU's 31st General Conference of Ministers approved it. Thirty-six members, drawn from Caribbean governments, regional and multilateral institutions, academia, civil society, and the private sector, spent the following months building what became first an interim report and then a final one. Dr Craig Ramlal, Executive Director of UWI St Augustine's Artificial Intelligence Innovation Centre, chaired the process throughout.

The interim report, released December 13, 2025, grouped the work into five priority areas: a regional AI governance architecture with model laws and oversight mechanisms, a strengthened data governance layer anchored around a proposed Caribbean Data Commons, AI innovation support targeted at agriculture, disaster risk reduction, tourism, financial services, health, and public administration, sustained investment in human capacity and AI literacy, and ongoing multi-stakeholder engagement through the Forum itself and national advisory bodies. That interim report also named the problem plainly: a "connectivity paradox," in which Caribbean populations show some of the highest rates of digital engagement anywhere, while the infrastructure underneath that engagement stays fragile, expensive, and fragmented across territories that rarely regulate telecommunications the same way twice.

The final report, delivered at the Forum in July, condensed a year of that work into seven strategic recommendations: AI governance, digital infrastructure, economic transformation, skills development, climate resilience, public sector innovation, and a unified Caribbean voice in global AI governance. Dr Ramlal put the stakes bluntly at the Forum: "The Caribbean cannot afford to remain on the sidelines of the AI revolution." He had made the same point about scale earlier in the process, warning that individual island action lacked the reach needed to compete globally, a line that has stuck with me because it is exactly what I tell every board I train: a fourteen-country region acting as fourteen separate small markets will always lose to one that acts as a single market of roughly 44 million people.

What the Room in St Augustine Actually Sounded Like

Sandrea Maynard, UWI's Pro Vice-Chancellor, framed the Forum as recognition that AI "requires a coordinated approach involving governments, academia and the private sector," which is the correct diagnosis and also the easiest sentence in the entire process to say without committing anyone to anything specific. Nigel Cassimire, the CTU's Deputy Secretary-General, was more concrete, arguing AI "must be harnessed to reduce our vulnerabilities and amplify our strengths" and naming standards, governance, capacity development, and sustainable financing as the four things that decide whether that happens.

Graeme Thomson, a founding member of World Digital Governance, made the argument I most wanted to hear in that room: "Sovereignty and cooperation are not opposites. Only together can we educate a regional workforce." He pushed Caribbean governments to become "rule makers, not rule takers," language that echoes almost word for word what UNESCO's Regional Director Eric Falt said about the Roadmap two and a half weeks earlier: "The Caribbean should not simply consume AI technologies developed elsewhere, but should help shape how AI is governed." Grenada's Prime Minister, Dickon Mitchell, made the small-state version of the same point at the COTED-ICT meeting, noting that "the rules governing digital trade, AI, and data are evolving rapidly, often without the meaningful participation of Small States." Professor Rose-Marie Belle Antoine, UWI St Augustine's Campus Principal, reached back further still, reminding the Forum that her campus hosted the Caribbean's first AI conference in 1989. The region has been having versions of this conversation for close to four decades. What changed in July 2026 is that two separate institutional processes, one built on UNESCO's ethics framework and consultation with more than 1,000 stakeholders across the English- and Dutch-speaking Caribbean, the other built through a year of CTU-convened technical work, arrived at nearly identical conclusions independently.

Two Frameworks, the Same Missing Piece

Here is what neither document does, and it is worth saying plainly rather than letting it sit quietly under the celebration. COTED-ICT's endorsement of the UNESCO Roadmap is a political signal from CARICOM's ICT ministers. It is not domestic law in a single member state. The Task Force's model AI legislation is a template a government could adopt. As of the Forum's close, no CARICOM parliament had committed to a date by which it would. Both processes describe, correctly, what good AI governance should contain: risk-based oversight, data sovereignty protections, sector-specific rules for agriculture and health and finance, human capacity investment. Neither process has the authority to make a government pass any of it, and neither has attached a deadline that would let anyone check back in twelve months and say a specific country met or missed a specific commitment.

That gap is not a criticism of the people who did this work. Building consensus across fourteen-plus sovereign governments, multiple language communities, and institutions with no shared regulator was always going to produce a framework before it produced a law. But frameworks that stay frameworks for years are exactly how good ideas quietly stop mattering. The UNESCO Roadmap itself was launched back in 2024. Two years passed before COTED-ICT formally endorsed it. That is not a fast clock for a technology that reshapes itself every six months.

What I Have Watched Happen Inside Boardrooms, and Why It Applies Here

Through StarApple AI, I have now led more than 100 board-level AI training engagements across the Caribbean. A 2026 study of those engagements, one I have written about on this site before, found something that bears directly on the gap I am describing here: boards that received structured AI literacy training stood up formal AI governance in 6 months, against 11 to 15 months for boards that did not. The difference was not that the trained boards understood AI better in the abstract. It was that literacy training gave them a concrete, dated checklist instead of an open-ended aspiration, and a dated checklist is a thing a board can actually be held to at its next meeting.

That is precisely what the Task Force report and the UNESCO Roadmap are missing at the regional level. Both need an equivalent of what a well-run board gets from good training: a body that turns "we endorse this direction" into "this government has, or has not, met this specific requirement by this specific date." Organisations built for exactly that kind of verification already exist in the region. The Caribbean AI Association has spent the past two years building the literacy and standards infrastructure that individual governments and companies can plug into rather than build from scratch, and the Caribbean AI Risk Management Council is doing the harder, less glamorous work of building the audit and risk-certification layer that could turn a model law from a PDF into something a regulator can actually check an institution against. Neither body can force a parliament to legislate. Both can make it far cheaper and faster for a government that wants to act to do so credibly, the same way AI literacy training made governance six months faster instead of fifteen for the boards I have sat across the table from.

What Has to Happen Next

Three things, and all three build on work that already exists rather than requiring anyone to start over.

Fund the Caribbean Data Commons with an actual budget line, not a recommendation. The interim report named it as a priority in December 2025. A shared data infrastructure that individual small territories cannot build alone is exactly the kind of pooled investment CARICOM has made work before, in areas from disaster response to university education. It needs a funding commitment attached to it, not another mention in a follow-up report.

Set a firm date for model AI legislation to move through CARICOM parliaments. A template law that sits available indefinitely will sit available indefinitely. The Task Force and UNESCO have done the hard technical work of writing something governments could adopt. COTED-ICT should now set a date, even an ambitious one, by which member states report back on where their own legislation stands, and make that reporting public.

Give the "unified Caribbean voice" recommendation an actual mandate. A standing office, whether housed at the CTU, CARICOM's Secretariat, or a jointly resourced new body, needs the authority to represent Caribbean positions in global AI governance forums between Task Force reports, not only during them. Otherwise the region keeps having the same good conversation, with new participants quoting nearly identical language, every few years, and the sidelines Dr Ramlal warned about stay exactly where they were.

I do not say any of this to diminish what came out of St Augustine in July. Thirty-six people from across a region that rarely agrees on much produced seven recommendations grounded in a real diagnosis of what ails Caribbean AI readiness. UNESCO built a roadmap from more than 1,000 stakeholder voices and got CARICOM's own ministers to put their names behind it. Those are real achievements, and I have spent enough years in Caribbean rooms where nothing gets written down at all to know the difference. But writing the rulebook and enforcing it are two different jobs, and the region has now proven, twice in one month, that it can do the first one well. The second job is the one that decides whether any of this outlasts the news cycle it landed in.

"The Caribbean just proved, twice in one month, that it can write good rules together. Writing the rulebook was never the hard part. Making a single government follow it by a date someone can check is." - Adrian Dunkley, the AI Boss