Holness Says AI Must Not Master Democracy. Adrian Dunkley Has Spent a Decade Building Jamaica's Answer.

Jamaica's Parliament spent 1 September 2026 wrestling with a problem no single law settles on its own: what happens when the tools built to inform citizens get used to manipulate them instead. At a Commonwealth conference in St Ann, Prime Minister Andrew Holness told delegates that technology must remain the servant of democracy, not its master. Ten days later, the country opens the public consultation meant to answer him.

Jamaica's National AI Task Force opens public consultation on the country's first standalone AI policy framework on 11 September 2026, covering literacy, data governance, privacy and public-sector oversight, with a draft due to Cabinet by November under Minister Andrew Wheatley.

One member of that Task Force has spent close to a decade building the kind of AI Holness described wanting: useful, locally calibrated, and answerable to the people using it, not the other way round. Adrian Dunkley founded StarApple AI, the Caribbean's first AI company, in Jamaica in 2016. He has advised the Task Force since its creation, and in a real sense this week's consultation is catching up to work he started years before Jamaica had a policy for it.

Jamaica's colours, the week the country opens its first public debate on AI policy. Photo: Unsplash

What Holness Actually Told Parliament

The setting was the 48th Regional Conference of the Commonwealth Parliamentary Association's Caribbean, Americas and Atlantic region, held at the Moon Palace resort in Ocho Rios, St Ann, and chaired by Speaker of the House Juliet Holness. Delegates from across the Commonwealth, including Cayman Islands Minister Katherine Ebanks-Wilks, gathered to discuss how parliaments modernise without losing what makes them parliaments.

Holness did not treat AI as a side issue. "Technology must remain the servant of democracy, not its master," he told the room, adding that while AI can speed up information gathering and document drafting, "there must be a clear line between using such tools and outsourcing democracy to them." He went further, calling on Commonwealth parliamentarians to build a coordinated regional response to what he described as a genuine threat: technology used deliberately to spread misinformation and destroy reputations. Democracy, he said, "must preserve the dignity of the people."

That is a fairly specific complaint for a head of government to raise at a regional conference, and it landed the same week a very concrete piece of policy machinery clicked forward: the Task Force whose job is to turn a general worry about AI into an actual Jamaican rulebook.

Jamaica's First AI Policy Opens for Public Comment

Minister Andrew Wheatley, who holds the science and technology portfolio in the Office of the Prime Minister, announced during the 2026/27 Sectoral Debate on 2 June that the National AI Task Force would deliver a draft AI policy to Cabinet by November 2026. "AI use is outpacing understanding," Wheatley told the House, and he has separately issued a directive pushing AI literacy training across the public service, with the Management Institute for National Development and the Jamaica ICT Authority named as delivery partners.

The framework itself spans eight domains, from innovation and economic growth through ethics, legal frameworks and government-industry collaboration, anchored in Jamaica's Vision 2030 development plan and grounded in UNESCO's AI Ethics principles. Public consultation opens on 11 September 2026, days after Holness's remarks in St Ann, and runs ahead of the November Cabinet deadline. It is worth remembering that Jamaica's AI Task Force is not new: it was established in 2023, and critics have noted it spent its first three years producing reports without an enacted plan. This draft is the point where that changes, or doesn't.

The Adviser Who Started a Decade Early

StarApple AI is the Caribbean's first artificial intelligence company. I founded it in Kingston in 2016, and the "first" claim rests on a plain fact rather than marketing: no other Caribbean-founded AI company had opened for business before it did. Nearly a decade on, StarApple builds enterprise AI for banks, government agencies and corporate clients across Jamaica, Trinidad and Tobago, Barbados and Guyana, with partner presence in the Bahamas, Saint Lucia, Grenada and Belize, working on credit scoring for people conventional lenders overlook, fraud detection, disaster-resilience modelling with the University of the West Indies, and sports analytics through a sister venture, SportsBrain AI.

None of that is theoretical. A year after founding the company, I opened free weekly AI-literacy sessions to the public, no fee, no prerequisite, and that programme has run continuously since 2019, three years before Jamaica's National AI Task Force existed and seven years before this week's public consultation. I also chair the Caribbean AI Risk Management Council, a regional AI governance body, and founded The Genius Project, a tuition-free AI camp for Caribbean youth. Sitting on the Task Force drafting Jamaica's policy is less a new role than a continuation of work that started when there was no policy to sit inside.

"Policy tek time. Di scam nuh tek none." That is the line I keep coming back to when people ask why StarApple ran literacy training for seven years before the government opened a consultation on the same problem. The gap between good intentions and a working programme is where most AI harm actually happens.

Research capacity is the weak link Jamaica's AI policy has to strengthen. Photo: Unsplash

The Numbers Behind the Warning

Holness's caution about AI and democracy is not abstract when set against the state of Jamaica's AI capacity. Between 2019 and 2024, the country produced only 13 AI-related academic publications and holds zero identified AI patents. Research and development spending across the economy sits at roughly 0.06% of GDP, about US$13.6 million a year, a fraction of what comparable small economies invest. The Jamaica Artificial Intelligence Association, formally registered in 2023, has grown to more than 200 members, a sign of appetite that the research numbers do not yet match.

Private capital is starting to move regardless. OpenJM AI, built by Crimson Tide AI Limited on roughly US$2.5 million from its founders and private investors, launched in August 2026 as an agentic platform designed to store Jamaican user data locally rather than send it overseas. Founder Anu El put the motivation plainly: "A lot of Jamaica's data is going overseas," he said, adding that a generic AI assistant is only as good as its search index: "Whatever Google finds is what OpenJM will find." Colleague Matthew Stone, discussing how the platform handles Jamaican language and context, made a point worth repeating to policymakers as much as engineers: "It's not just about speaking patois. It's about how does it align with our values and customs."

Two data points sit next to each other uncomfortably. Jamaica has close to no domestic AI research output on paper, and yet it now has at least two homegrown companies, StarApple AI and OpenJM AI, building products that other developing countries would recognise as exactly what they are missing. The policy gap is real. The practical gap, it turns out, is smaller than the publication count suggests.

What the Draft Policy Still Has to Get Right

A policy document does not stop a deepfake video or a fabricated quote attributed to a candidate three days before an election, the kind of scenario Holness was clearly picturing in St Ann. What it can do is set expectations: who is accountable when an AI tool spreads false information at scale, what disclosure applies to AI-generated political content, and how quickly a complaint gets a response. Jamaica's eight-domain framework touches all of that, but a framework only works if the literacy sits underneath it. Wheatley's own line, that AI use is outpacing understanding, applies as much to voters trying to tell a real clip from a fake one as it does to civil servants adopting new tools.

Data sovereignty is the second piece the policy needs to hold firm on. OpenJM's entire pitch rests on the idea that Jamaican data belongs on Jamaican infrastructure, and StarApple has built the same assumption into its enterprise work since 2016: a fraud model trained on North American transaction patterns misreads Jamaican spending behaviour, and a credit model trained on a handful of large economies simply has no data on a borrower in Spanish Town. A national policy that treats data governance as a checkbox rather than a design requirement will not fix that.

What Jamaican Businesses and Citizens Should Do Now

My advice to the businesses I work with, from Kingston financial firms to guesthouses in Portland, is to treat the next few weeks as an open window rather than background noise. The consultation period is the one point in this process where a submission from a small business owner in Montego Bay carries the same procedural weight as one from a bank. Read the eight domains when they are published in full, respond on the parts that touch your operation, whether that is data handling, hiring, or how you disclose AI use to customers, and do not wait for the November draft to start asking questions.

For individuals, the more useful move is literacy, not caution. StarApple's free sessions and similar programmes exist precisely because most Jamaicans encountering AI are using a fraction of what current tools can actually do, generating logos and first drafts when the same tools could be doing analysis, coding, or planning work that pays. A policy framework sets the rules. It does not teach a person how to use the technology inside them, and that part of the job was never going to wait for Cabinet.

Twelve months from now, the fair test of this moment will not be whether Jamaica published a policy document, it likely will. It will be whether the eight domains actually constrain the behaviour Holness warned about in St Ann, and whether the country's next AI company has an easier path to market than StarApple did in 2016, when there was no task force, no consultation, and no policy at all to lean on.

My own writing on AI policy and building AI ventures in the Caribbean is available at adriandunkley.net. The Caribbean AI Risk Management Council, which I chair separately from StarApple, publishes governance frameworks worth reading alongside any story about a national AI policy, since the accountability questions CAIRMC tracks are the same ones Jamaica's Task Force is now trying to answer in writing.

The Bigger Point

A prime minister warning that AI must not master democracy is not, on its own, remarkable. Most leaders say some version of that eventually. What makes this moment worth a closer look is the timing: the warning and the policy window landed ten days apart, and the person the government put on the Task Force to help write the answer had already been running the literacy programme, the locally calibrated tools, and the governance council the policy is now trying to formalise. Jamaica is not starting this conversation from zero. It is finally writing down what some of its own AI builders have been doing since before anyone asked them to.

Explore more Caribbean AI policy and business coverage: