CARICOM's Trade Ministers Endorsed a Caribbean AI Roadmap on 7 July. It Carries No Binding Risk Standard for a Single AI System in the Region.

On 7 July 2026, the Council for Trade and Economic Development configured for Information and Communication Technologies, known across CARICOM as COTED-ICT, held its 126th Special Meeting and endorsed the UNESCO Caribbean Artificial Intelligence Policy Roadmap. UNESCO's Regional Director for the Caribbean, Eric Falt, described the moment as the region choosing to help shape how AI is governed rather than simply consuming systems built elsewhere. Grenada's Prime Minister, Dickon Mitchell, used the same platform to say something less celebratory: "The rules governing digital trade, AI, and data are evolving rapidly, often without meaningful Small States participation." Both statements were made at the same meeting, about the same document, and they do not sit easily together.

The Roadmap itself is not new. UNESCO first published it in 2021 and refreshed it through 2024, grounding it in the UNESCO Recommendation on the Ethics of AI and building it out of consultations with more than 1,000 stakeholders across the English- and Dutch-speaking Caribbean. What changed on 7 July was political status: fifteen CARICOM trade ministries collectively put their name to it. That is worth taking seriously. It is also worth being precise about what it does and does not commit any government to.

What COTED-ICT Actually Endorsed

The Roadmap organises itself around four pillars. Culture and Creativity addresses AI's effect on Caribbean cultural identity, language, and creative industries. Upskilling and Education covers workforce readiness and AI literacy. Resilience and Sustainability looks at AI's role in disaster preparedness and environmental management, an area with obvious relevance after a decade of intensifying hurricane seasons. The pillar that matters most for this article is Governance and Transformation, which the Roadmap describes as establishing a regulatory framework, defining roles and responsibilities, and creating an oversight mechanism so AI systems comply with ethical and legal standards.

Read that description again slowly. It commits the signatories to building a framework. It does not itself set out what the framework requires an AI system to pass before deployment, how often a deployed system gets re-tested, what counts as a reportable AI incident, or which agency a citizen can appeal to when an automated decision affects their benefits, their credit, or their liberty. UNESCO's own Readiness Assessment Methodology, which the organisation says has been completed or is underway in several Caribbean countries, is explicitly designed to identify policy gaps and institutional capacity gaps, which is a useful diagnostic tool but a different instrument from a risk standard with teeth.

None of this is a criticism of UNESCO's work, which fills a role no other body in the region has filled: a shared regional vocabulary for AI policy, built from genuine consultation rather than a template copied from Brussels or Washington. The criticism, if there is one, belongs to the gap between what got endorsed on 7 July and what a compliance officer, a hospital administrator, or a bank's risk committee actually needs on the Monday morning after an AI system produces a decision nobody can explain.

A Roadmap Answers a Different Question Than a Risk Standard

A regional policy roadmap asks where a group of governments wants AI investment and capacity to go over the next several years. A risk standard asks a narrower question: what has to be true before a specific AI system goes live today, and what happens the moment it fails. Those are separate disciplines, built by different kinds of institutions, and the seven months so far in 2026 have shown what each one produces when it moves.

The clearest recent example of the second kind sits inside the region already. The Centrale Bank van Curaçao en Sint Maarten published its Financial Stability Report 2026 in May, naming AI-generated deepfakes directly as a financial-stability risk and committing, on the record, to developing supervisory guidance covering cybersecurity, third-party technology providers, operational resilience, and AI across the banks, insurers, and pension funds it regulates. That is one paragraph in one annual report from a regulator overseeing one of the smaller financial systems in the region. It names a specific mechanism, attaches it to specific institutional categories, and comes from a body with actual supervisory power over the entities it is writing about. The UNESCO Roadmap, endorsed at a much higher political level and covering fifteen governments at once, does neither. Broader endorsement bought less operational specificity, not more, and that trade-off is exactly what a reader should expect from a document built to secure agreement across fifteen very different economies rather than to regulate one financial sector.

The Readiness Index the Region Would Rather Not Compare Against

The political win of 7 July lands against an inconvenient backdrop. Oxford Insights published its Government AI Readiness Index 2025 in the weeks before the COTED-ICT meeting, scoring national governments across six pillars: Policy Capacity, AI Infrastructure, Governance, Public Sector Adoption, Development and Diffusion, and Resilience. Jamaica is the region's strongest performer, sitting inside the global top 100, a result Oxford Insights attributes largely to the country's published national AI strategy. Trinidad and Tobago ranks 122nd globally. The Bahamas sits at 126th. Antigua and Barbuda, according to reporting in the Antiguan Herald on 30 April 2026, ranks 154th out of the countries assessed.

Governance is not a side note in that index. It is one of the six pillars a government is scored against directly, alongside the policy capacity and public sector adoption that determine whether a roadmap actually turns into practice inside a ministry or a regulator. A region can hold a unanimous regional endorsement ceremony and still score in the bottom half of a global governance ranking, because the two measure different things: one measures whether governments agreed on a direction, the other measures whether the machinery to execute that direction exists yet inside each government. As of July 2026, CARICOM has strong evidence of the first and weak evidence of the second.

Photo via Unsplash

Small States, Big Rules: What the Grenada Quote Actually Signals

Prime Minister Mitchell's comment about "meaningful Small States participation" was not offered as a complaint about the Roadmap itself. It was aimed at the wider set of digital trade, data, and AI rules being written in Brussels, Washington, and Geneva, rules Caribbean governments will eventually have to comply with whether or not they had a seat at the table when the rules were drafted. That framing matters for a risk governance body like CAIRMC for a specific reason: a region that already lacks a voice in shaping the external rules it will be held to cannot afford to also lack the internal capacity to enforce its own. Endorsing a shared roadmap is a reasonable first move toward the first problem. It does nothing, on its own, to fix the second.

The two gaps compound each other. If CARICOM states are not shaping the EU AI Act, the OECD's AI governance principles, or the emerging US state-level AI rules that will bind Caribbean exporters and financial institutions with cross-border exposure, the region needs stronger domestic risk capacity precisely to interpret and adapt those external rules quickly once they land. A low score on Oxford Insights' Governance pillar today means a slower adaptation the next time a rule written elsewhere takes effect on a date the region did not choose.

A Second Regional Process Is Closing Out the Same Month

Readers following Caribbean AI policy closely should keep two separate processes distinct, because the calendar has put them within two weeks of each other. COTED-ICT's 7 July endorsement belongs to the UNESCO Roadmap, coordinated through CARICOM's trade and ICT ministries. A different body, the Caribbean Telecommunications Union, launched its own Caribbean AI Task Force in July 2025 and is set to release the Task Force's final report at the inaugural Caribbean AI Forum, running 23 and 24 July at the University of the West Indies' St Augustine campus in Trinidad and Tobago. That Task Force's five priorities, regional governance, data sovereignty, innovation, human capacity, and multi-stakeholder engagement, overlap conceptually with the UNESCO Roadmap's four pillars without being the same document, the same convening body, or the same governance mechanism.

Two parallel regional AI policy efforts producing overlapping but distinct roadmaps inside the same month is not, by itself, a bad sign. It suggests genuine institutional appetite for the subject. It does raise a coordination question CARICOM has not yet answered publicly: which of the two processes, if either, is meant to convert into the binding risk standard that a bank, a hospital, or a government ministry running an AI system today actually needs. Until that question has an answer, the practical position for any Caribbean organisation deploying AI is unchanged regardless of which roadmap eventually wins the coordination argument.

What Actually Needs to Exist Before Endorsement Becomes Enforcement

Turning a roadmap's Governance and Transformation pillar into something a compliance officer can act on requires specific, testable commitments that neither of this month's regional documents currently contains. A defined pre-deployment testing requirement, stating what an AI system has to demonstrate before it can be used to make a decision that affects a person's finances, health, or legal status. An incident-reporting duty, with a threshold for what counts as a reportable AI failure and a named recipient for that report. A named accountable individual or office inside every institution deploying a high-risk AI system, so responsibility does not dissolve into "the vendor built it" when something goes wrong. A right of human review, so a citizen or customer affected by an automated decision has somewhere to appeal beyond the system that made it. And a public register of high-risk AI systems in government use, the kind several EU member states are now building under the AI Act, so oversight bodies and the public know what is actually running.

None of those five items require CARICOM to wait for a treaty or a new supranational institution. Any single member state, or any regulator inside one, could adopt all five for the sectors it supervises this year, independent of whichever regional roadmap eventually becomes the reference document. Jamaica and Trinidad and Tobago, the two states furthest along on their own national AI strategy work, are the most plausible candidates to move first, and a first mover here carries the same reputational advantage CBCS captured by publishing before larger regulators in the region.

Where CAIRMC Fits Into This Gap

The Caribbean AI Risk Management Council exists precisely because the interval between a regional roadmap and a binding national standard tends to run into years, and organisations running AI systems today do not have years to wait. CAIRMC's Caribbean AI Risk Assessment methodology, CARA, gives an institution a structured way to classify an AI system by risk tier and attach proportionate controls, whether the system was built in-house or inherited through a vendor contract, without needing a finished piece of legislation to reference first. The Qualified AI Risk Professional certification CAIRMC administers gives a compliance function the specific competency an AI audit or an incident investigation requires, built for Caribbean institutional conditions rather than adapted from a generic global template.

Adrian Dunkley, who chairs CAIRMC and founded StarApple AI, recognised as the first artificial intelligence company built in the Caribbean, has made a consistent argument across the past two years: policy roadmaps and risk governance are separate disciplines that need to run at the same time, not one after the other. The organisations most exposed to AI risk today, banks running credit models, hospitals running triage tools, government agencies running eligibility systems, cannot make that risk wait on a regional coordination process to finish deciding which roadmap takes precedence. His writing, and CAIRMC's published frameworks, are available through starappleai.org and adriandunkley.net. Sister organisations tracking national-level detail behind this month's regional headlines, including AI Trinidad and Tobago and the wider Caribbean AI Association, are documenting exactly the adoption patterns any eventual risk standard will need to be built around rather than guessed at.

What This Leaves Unanswered

Whether the UNESCO Roadmap or the CTU's Task Force process ends up producing CARICOM's first binding AI risk instrument, or whether the two get merged into something neither body currently plans, is a genuinely open question this article cannot resolve, because CARICOM itself has not said which track, if either, is meant to carry that weight. What is not open to much doubt is the readiness gap Oxford Insights measured before either document was signed or launched. A region can hold a unanimous endorsement ceremony and still rank in the bottom half of a global governance index the same month, and both facts were true on 7 July 2026. The Caribbean AI Forum, opening in Trinidad and Tobago the day after this article publishes, is the next moment CARICOM has to start closing that gap rather than widening it with a third overlapping roadmap.

Frequently Asked Questions

What did COTED-ICT actually endorse on 7 July 2026?

COTED-ICT, the CARICOM organ handling trade and information and communication technology policy, endorsed the UNESCO Caribbean Artificial Intelligence Policy Roadmap at its 126th Special Meeting. The Roadmap, first published in 2021 and refreshed through 2024, sets out four pillars: Culture and Creativity, Governance and Transformation, Upskilling and Education, and Resilience and Sustainability. The endorsement is a political and diplomatic act, not the adoption of binding legislation.

Does the UNESCO Roadmap set binding AI risk standards?

No. The Governance and Transformation pillar commits signatories to building a regulatory framework, defining roles and responsibilities, and creating an oversight mechanism. It does not itself specify a testing requirement, an incident-reporting duty, an audit cadence, or an accountability chain for a specific deployed AI system. Those remain to be built by individual member states or sector regulators.

How is the UNESCO Roadmap different from the CTU's Caribbean AI Task Force?

They are separate, parallel initiatives. The UNESCO Roadmap is coordinated through CARICOM's trade and ICT ministries via COTED-ICT. The Caribbean AI Task Force was launched in July 2025 by the Caribbean Telecommunications Union and is releasing its final report, with five priorities of its own, at the Caribbean AI Forum on 23 and 24 July 2026 in Trinidad and Tobago. The two efforts overlap in subject matter without being the same document or governance mechanism.

What does the Oxford Insights Government AI Readiness Index 2025 say about the Caribbean?

Jamaica is the region's strongest performer and the only CARICOM state inside the global top 100, a result Oxford Insights attributes to the country's published national AI strategy. Trinidad and Tobago ranks 122nd globally, The Bahamas 126th, and Antigua and Barbuda 154th, according to reporting in the Antiguan Herald on 30 April 2026. The index scores governments across six pillars, including one specifically for Governance.

What did Grenada's Prime Minister say at the COTED-ICT meeting?

Dickon Mitchell said that the rules governing digital trade, AI, and data are evolving rapidly, often without meaningful participation from Small States. The comment was aimed at global AI and digital trade rulemaking beyond the Caribbean's own roadmap, underscoring that the region faces a capacity gap both in shaping external rules and in enforcing its own.

What would it take to turn the Roadmap's governance pillar into an enforceable standard?

Five concrete elements: a defined pre-deployment testing requirement, an incident-reporting duty with a clear threshold, a named accountable officer inside every institution running a high-risk AI system, a right of human review for people affected by automated decisions, and a public register of high-risk government AI systems. None of these require CARICOM to wait for a new treaty; any member state or sector regulator could adopt them independently.

What should a Caribbean organisation do while the region sorts out which roadmap leads?

Build a working risk framework now rather than waiting for either regional process to convert into binding law. That means classifying AI systems by risk tier, documenting who is accountable for each one, and establishing an incident response plan, whether the system was built in-house or supplied by a vendor. CAIRMC's CARA methodology and QAIRP certification are built for organisations in exactly this position.

Does CAIRMC compete with UNESCO or the CTU on this?

No. UNESCO and the CTU each convene governments and set policy direction at a regional level. CAIRMC works directly with the banks, insurers, hospitals, and government agencies deploying AI systems today, publishing sector risk guidance and running a certification programme. The region needs the policy-level work and the operational risk work running at the same time, not in sequence.

Related reading across the Caribbean AI network

This article sits alongside ongoing coverage of AI governance, risk, and company-building across the region. For related perspectives:

• Adrian Dunkley's own analysis of Caribbean AI governance and risk leadership
• StarApple AI, the company building Caribbean-specific AI products and infrastructure
• AI Trinidad and Tobago, tracking policy developments in the state hosting this month's Caribbean AI Forum
• Caribbean AI Association for regional adoption trends across the sectors this governance gap affects