
Regulation & Compliance / AI risk & governance
Brantley’s AI Protocol Proposal Needs Clear Local Responsibilities
Nevis’s Premier proposes common Caribbean AI protocols. Their value depends on clear authority, current legal checks and practical safeguards for each use.
Nevis’s Premier, Mark Brantley, has proposed a dedicated CARICOM meeting on artificial intelligence and common regional protocols. The useful next question is who would turn those proposals into decisions that a ministry, school or small business can follow.
Shared work could reduce duplicated drafting and help small administrations ask suppliers better questions. It still needs a clear legal basis, staff who can apply the rules and a way for affected people to challenge mistakes. An office should be able to identify who is responsible for its AI use while regional discussions continue.
TL;DR
- Brantley’s September 2026 proposal covers nine areas, including data, public procurement, discrimination and human oversight. It does not itself create binding obligations.
- The current commencement status of the Federation’s 2018 Data Protection Act was not established in this review. A dated secondary source cannot settle it.
- Organisations can examine their existing AI uses, supplier terms and responsibilities now. A completed checklist is not a legal compliance certificate.
The Proposal and the Work Already Underway
In an op-ed published on 27 September 2026, Brantley called for regional leaders to bring educators, workers, businesses, researchers and technology companies into the discussion. His proposed topics include responsible AI use; personal data; cybersecurity; intellectual property; procurement; education and healthcare; discrimination; deepfakes; and transparency and human oversight. He also argued for public education and support for Caribbean innovators.
There is an existing regional policy process to build on. UNESCO’s 7 July 2026 announcement reported COTED-ICT’s endorsement of the Caribbean AI Policy Roadmap. Launched in 2024, the roadmap drew on consultations with more than 1,000 stakeholders in the English- and Dutch-speaking Caribbean. Endorsement provides a basis for coordination; it does not establish that every participating country has enacted the same law.
Cullen International’s September 2026 public summary reported that none of the 13 countries in its benchmark had adopted an AI-specific regulatory framework. Its coverage includes St Kitts and Nevis. That is a finding about the benchmark’s scope, not evidence that existing laws cease to apply when a system uses AI. We reviewed the public summary, not the full subscriber benchmark.
Check the Legal Position Before Building an Argument on It
The official Data Protection Act, 2018 records assent on 23 May and publication in the Gazette on 7 June 2018. Section 1 provides for commencement by a ministerial order published in the Gazette. Those dates establish enactment and publication, not the later commencement date.
A UWI resource says the Act had not been brought into force as of February 2025. That statement is explicitly dated. This review has not established the complete current position, and it would be unsafe to turn an unsuccessful search for a later order into a claim that the Act remains uncommenced.
Before relying on a particular duty, organisations should obtain current advice on the applicable legislation, amendments, commencement instruments and sector requirements. The Ministry of Justice and Legal Affairs is an official starting point for government legal information. This article does not establish the availability or absence of a particular complaint route.
Consider a hypothetical guesthouse employee using a chatbot to draft a booking reply. Uploading a passport scan would create a different exposure from asking for help with a generic greeting. The employer should assess what information is genuinely needed, which tool is approved and what the supplier may retain or disclose. These questions remain useful regardless of the outcome of a commencement check. They do not establish that every upload is lawful or that a guest has no legal protection.
What Regional Cooperation Could Make Easier
World Bank indicators put the Federation’s population at 46,922 in 2025 and internet use at 77% in 2024. Shared technical expertise may help a small administration, but population alone does not establish what regulatory structure it can afford or should choose.
One practical option is a common supplier questionnaire, adapted to local law and the intended service. It could ask where information is processed, which subcontractors can access it, how retention works, whether inputs may train models and what happens when the contract ends. Answers should be checked against the actual service configuration and contractual commitments. A supplier’s prompt response is useful evidence to examine, not proof of safety.
Another option is to share test methods and procurement expertise while keeping responsibility explicit. A regional template should identify which body adopts it, which organisations it covers and how a person can seek review. For the Federation, implementation must account for the respective powers of federal and Nevis institutions rather than assuming a regional statement settles that allocation.
Participation Matters, but One Meeting Is Not the Whole Record
An IDB release dated 21 September 2026 listed officials from twelve governments at an AI meeting alongside technology executives. St Kitts and Nevis was not named. The release does not explain the invitation process, and the omission does not prove exclusion from AI policy discussions more broadly.
Adrian Dunkley’s Preparation Asymmetry describes a structural disadvantage for countries using AI systems designed elsewhere. Applied here, it suggests asking how local institutions and affected communities can influence requirements, testing and supplier changes. CAIRMC’s application of that idea is an editorial recommendation. It is not a finding that a particular vendor or government has denied anyone a voice.
Participation should extend beyond negotiating access to a product. Teachers, workers and service users can help identify failures that a generic demonstration misses. Their input needs a route into procurement and oversight decisions, with enough information to make that input meaningful.
Plan for Changes in Work Without Treating Exposure as Job Loss
A July 2024 ILO–World Bank study estimated that 26–38% of jobs across Latin America and the Caribbean could be exposed to generative AI. It placed 2–5% at risk of full automation under the capabilities assessed, while 8–14% could gain productivity. These regional estimates describe different kinds of potential exposure. They are not a forecast for St Kitts and Nevis or proof of losses already occurring.
For a local employer, a more useful starting point is the actual work. Which tasks might change? What skills would staff need? Who would review errors, and would that work fit within their paid time? Worker consultation and a realistic training plan should inform a deployment decision, rather than follow it as an afterthought.
A Practical Review for an Organisation
- List the uses that matter. Record the purpose, responsible person, affected people and information involved. Begin with sensitive or consequential activities.
- Separate legal questions from policy choices. Record the current advice and source for each relevant obligation. Label unresolved questions rather than scoring them as proof that no law applies.
- Check the service in use. Compare supplier promises with settings, contracts, access controls and retention arrangements. Keep unnecessary personal data out of exploratory tests.
- Give staff a usable policy. Explain approved tools, information restrictions, human review and where to ask for help. Provide a workable alternative when a tool is unsuitable.
- Assign follow-through. Set deadlines that reflect the risk and available staff, document the decision and revisit it when the system or applicable requirements change.
This review can reveal unanswered questions. It cannot replace legal advice, a sector-specific assessment or evidence that safeguards work in practice. Regional protocols would be most useful if they helped institutions carry out that work consistently and publish who is accountable.
Frequently Asked Questions
What did Mark Brantley propose?
A dedicated CARICOM Heads of Government meeting on AI, with common regional principles and protocols. It is a political proposal, not an enacted regional law.
Is the St Kitts and Nevis Data Protection Act in force?
This review has not established its current commencement status. The 2018 text requires a ministerial Gazette order. UWI’s statement that it was not in force refers to February 2025 and cannot settle the position today.
Does uncertainty about that Act mean personal data has no protection?
No. Other applicable legal, contractual and professional duties must be considered. An incomplete commencement search cannot establish that no remedies or duties exist.
Do regional AI job estimates predict losses in St Kitts and Nevis?
No. Exposure to a technology differs from a forecast of job losses. Regional estimates do not establish outcomes for a particular island, employer or occupation.
What can an organisation do while regional work continues?
Identify its AI uses, obtain current legal advice where needed, document supplier data practices and give staff approved tools and a clear route for questions or incidents. Prioritise consequential uses rather than treating a checklist score as proof of compliance.
Publisher Disclosure
CAIRMC is chaired by Adrian Dunkley, founder of StarApple AI and president of the Caribbean AI Association. This article offers governance analysis of public sources, not legal advice on a particular organisation.
CAIRMC is a council of the Caribbean AI Association. Browse more research · Browse by topic