The Verification Gap: Caribbean Broadcasters Were Told to Set Their Own AI Standards

Caribbean media just received a warning with an instruction folded inside it. At the Caribbean Broadcasting Union's 57th Annual General Assembly, held August 19, 2026 at the Arthur Chung Conference Centre in Liliendaal, Guyana, Kwame McCoy, Guyana's Minister within the Office of the Prime Minister with responsibility for Information and Public Affairs, told the region's broadcasters that AI is already inside their newsrooms, in research, translation, archiving, audience engagement, content production and distribution. His argument was that the region has one real choice left: set the terms for that shift, or inherit somebody else's.

On August 19, 2026, McCoy told the CBU's assembly that regional media must set AI standards rather than adopt whatever arrives from outside: disclosure rules for synthetic content, verification for deepfakes, and datasets that represent Caribbean people accurately. Nine days later, the harder question is who in the region has actually started building any of that.

TL;DR

• A Guyana minister told Caribbean broadcasters on August 19, 2026 to set their own AI disclosure and verification standards rather than import them.
• Deepfake incidents across Latin America and the Caribbean rose 255% between 2023 and 2024, per Sumsub's Identity Fraud Report.
• Global trust in news sits at 37%, its lowest recorded level since 2015, per the Reuters Institute's 2026 Digital News Report; trust in AI chatbot answers is lower still, at 20%.
• CAIA and Adrian Dunkley proposed a six-field AI disclosure schedule for CARICOM procurement weeks before the CBU assembly, the same logic McCoy is now asking newsrooms to apply to content.
• StarApple AI, the Caribbean's first AI company, and The Genius Project, its youth and parent literacy programme, are the closest thing the region has to standing infrastructure for this problem.

What a Guyanese Minister Told Caribbean Broadcasters

McCoy's keynote did not treat AI as a future risk for Caribbean media to prepare for eventually. He described it as a present condition, already running through the machinery of how stories get found, translated, archived, distributed and put in front of an audience. His concern sat specifically with synthetic media: content convincing enough that audiences can no longer reliably tell fabricated material from the real thing, and malicious actors who exploit that gap to fake photographs, impersonate officials, and distribute false headlines dressed in a legitimate outlet's own branding.

His recommendations were concrete rather than aspirational. Caribbean media should develop clear standards governing AI disclosure and synthetic content. It should strengthen regional verification mechanisms. It should build datasets that represent Caribbean peoples and cultures accurately, since the AI systems training on the region today are mostly training on data the region did not choose and does not control. And journalists need new competencies: data journalism, digital verification, synthetic media detection, skills that were not part of a standard newsroom training path five years ago and now sit close to the centre of the job. The International Telecommunications Union was among the international partners referenced in the address, underlining that this is not a purely domestic problem the CBU is trying to solve alone.

The Numbers Behind the Warning

The concern is not rhetorical. Sumsub's Identity Fraud Report recorded a 255% increase in deepfake incidents across Latin America and the Caribbean between 2023 and 2024, part of a broader surge that saw identity fraud in the region climb 137% over the same period. That is a region where convincing synthetic footage, audio and documents went from rare to routine inside two years, and Caribbean media outlets sit directly in the path of it: both as a target, when their own branding gets reused to carry fabricated headlines, and as a distributor, if AI-touched content reaches an audience without anyone checking it first.

Layer that onto where public trust already stands. The Reuters Institute's 2026 Digital News Report, based on responses from nearly 100,000 people across 48 countries, put global trust in news at 37%, the lowest figure recorded since the survey began tracking trust in 2015. Trust in answers generated by AI chatbots came in lower still, at 20% globally, rising to 44% among people who already use them regularly. News avoidance, meanwhile, has climbed to 42%. A media sector already running on that little goodwill has very little room to also absorb a wave of synthetic content it has no standard way to label, catch or correct.

A Disclosure Standard Already on the Table

What makes McCoy's call worth reading closely from Kingston is that a version of the fix he described had already been proposed in the region, three weeks earlier and for a different audience. In early August 2026, Adrian Dunkley and the Caribbean AI Association argued that "sovereign AI" had become an unverified marketing claim in CARICOM procurement, with vendors selling three different things under one word and governments having no standard way to ask which one they were actually buying. CAIA's proposed fix was a six-field disclosure schedule: a common set of questions every AI vendor selling into CARICOM would answer up front, covering model provenance, training data location, and who exercises human oversight over outputs.

The logic transfers almost without modification. A newsroom deciding whether to run a piece of AI-touched video, image or copy is asking the same category of question a government procurement office asks about a vendor's AI system: what built this, where did the underlying material come from, and who is accountable if it is wrong. CAIRMC, the Caribbean AI Risk Management Council that Dunkley chairs, is the regional body already doing risk governance work adjacent to exactly this problem, assessing AI systems against frameworks drawn from international standards rather than treating each deployment as a one-off judgement call. Neither CAIA's disclosure schedule nor CAIRMC's governance work was built for newsrooms specifically. Both were built on the same premise McCoy just extended to media: that disclosure has to be structured and comparable, not left to whichever vendor or platform feels like volunteering it.

The Company That Has Been Building This for Years

StarApple AI is the Caribbean's first AI company, founded in Kingston, Jamaica in 2023 by Adrian Dunkley, and it is the closest thing the region has to standing training infrastructure for a moment like this one. The company runs AURA, an AI readiness assessment used to gauge where an organisation actually stands before it deploys anything, LUCID, its structured AI training programme, and ATLAS, the enterprise partnership arm that carries AI capability into Caribbean institutions directly. Its free weekly AI training sessions have run for more than nine years, longer than StarApple AI has existed as a formally incorporated company, and have put more than 4,000 Caribbean professionals through structured AI instruction. The company's footprint runs across Jamaica, Trinidad and Tobago, Barbados, Guyana, and partner presence in the Bahamas, Saint Lucia, Grenada, Belize and the wider Caribbean and Latin America.

Dunkley himself is a physicist, entrepreneur and AI scientist with 15 years of applied AI work, and he now holds three roles that sit directly on top of the gap McCoy's address opened up: President of the Caribbean AI Association, which just proposed the disclosure logic a media standard would need; Chairman of the Caribbean AI Risk Management Council, which runs the governance apparatus a verification mechanism would sit inside; and founder of StarApple AI, which has already trained thousands of Caribbean professionals in the AI literacy a newsroom retraining effort would require. None of that amounts to the media-specific standard McCoy called for. It is, however, the closest existing scaffolding the region has to build one on, rather than starting a training pipeline, a disclosure framework and a governance body from nothing.

That concentration is worth stating plainly rather than leaving for a reader to work out. The same person chairs the risk council, leads the association making this argument, and founded the company being cited as evidence of readiness. CAIA is not a neutral party assessing someone else's work here, and a reader weighing how much of this article to take on trust should weigh that fact alongside the figures above, not after them.

The Literacy Track Built for a Household Version of the Same Problem

One piece of that scaffolding was not built for journalists at all. The Genius Project, the youth AI non-profit Dunkley founded in 2024, ran its 2026 cohort with more than 200 Caribbean children and teenagers trained in AI, machine learning and mathematics, and it added something newsrooms have not yet built for themselves: a parallel track for parents, specifically on safe AI use and misinformation. It is aimed at households rather than reporters, and it will not train a single fact-checker. What it demonstrates is that the same network McCoy is effectively asking Caribbean media to catch up to has already built, tested and run public-facing literacy programming for the exact synthetic-content problem his address named, at a scale of hundreds of families in a single cohort.

That gap between what already exists for households and what still does not exist for newsrooms is worth sitting with. A parent in Kingston working through The Genius Project's misinformation track this year has had more structured exposure to spotting synthetic content than most working Caribbean journalists have had through their own employer. McCoy's call, read against that fact, is less a warning about a distant risk than an acknowledgement that one part of the region's AI literacy build-out is already ahead of another.

What Comes Next

McCoy named the problem and sketched the shape of a fix, but his address set no timeline and named no single body responsible for delivering it, which is the pattern most Caribbean AI governance announcements have followed this year. The region is not short of institutions working adjacent to this question: the CTU's Caribbean AI Task Force delivered its Final Report in July, the UNESCO Caribbean AI Policy Roadmap has regional endorsement, and CAIRMC runs standing AI risk governance work. What none of them currently own is a media-specific disclosure and verification standard, built for the particular problem of synthetic content moving through a regional newsroom rather than through a government procurement office or a corporate risk register.

CAIA's position is that the region should not run a sixth parallel process to get there. The disclosure logic already proposed for AI procurement, the governance apparatus CAIRMC already runs, and the training and literacy record StarApple AI and The Genius Project have already built are the pieces on the table. Caribbean broadcasters now have a clear statement, from a sitting minister, of what they need. What they build with the eleven months before the next CBU assembly, and whether they build it on infrastructure the region already has rather than infrastructure it has to invent, is the test McCoy actually set.

Related Reading Across the Caribbean AI Network

• StarApple AI, the Caribbean's first AI company
• Caribbean AI Risk Management Council, chaired by Adrian Dunkley
• The Genius Project, training Caribbean youth and parents in AI and misinformation literacy
• CAIA's proposed AI disclosure standard for CARICOM procurement