Trinidad Deepfaked a Finance Minister. Caribbean Insurance Claims Could Be Next.

The Fake Minister, Six Times Over

The pattern started in August 2025, when Trinidadian authorities began investigating scam videos impersonating the country's sitting Finance Minister, used to promote a fraudulent investment scheme. Cabinet responded within weeks, creating an Inter-Ministerial Steering Committee on Cyber Security and AI on 11 September 2025. It did not stop the pattern. By October 2025, the Senate had identified fabricated videos of former President Anthony Carmona and former Republic Bank chairman Dr Ronald Ramkissoon, both used for the same style of fake endorsement. The Trinidad and Tobago Securities and Exchange Commission (TTSEC) issued its sixth public investor alert on 26 June 2026, this time naming fabricated content impersonating former Prime Minister Dr Keith Rowley and Guardian Media editor Kejan Haynes.

Six alerts in under a year, against four separate public figures, is not a one-off incident. It is a functioning fraud pipeline that has already found an audience willing to trust a familiar face on video, even when that face never said the words attached to it. And as CAIRMC has pointed out in its own review of the pattern, a deepfake made in Port of Spain reaches Kingston, Bridgetown, or Georgetown on WhatsApp just as fast as it reaches a Trinidadian inbox. There is no border checkpoint for a video file.

What the New AI Roadmap Leaves Out

CARICOM's Council for Trade and Economic Development formally endorsed the UNESCO Caribbean Artificial Intelligence Policy Roadmap on 7 July 2026, eleven days after TTSEC's sixth alert. The roadmap is built around four pillars: Culture and Creativity, Governance and Transformation, Education and Upskilling, and Resiliency and Sustainability. It is the first document of its kind to win regional endorsement, and it is a genuine step forward for a region where, until recently, AI policy existed mostly as a talking point at conferences rather than a signed commitment.

It also has a specific hole. A review published by the Caribbean AI Risk Management Council found that the roadmap does not name deepfake-enabled financial fraud, investment scam impersonation, or AI-generated fake endorsements as distinct risk categories, despite the Trinidad wave running concurrently with the document's drafting. Nor does it establish a binding, shared, cross-border enforcement mechanism, meaning a fraud pattern that moves between territories in hours still has to be handled territory by territory, with no CARICOM-wide authority to act on it directly.

That is not a criticism of the roadmap's ambition. Broad governance documents rarely name specific fraud typologies, and this one covers education, culture, and infrastructure alongside risk. But an insurer reading it for guidance on AI-enabled claims fraud will not find any, which leaves the sector's own frameworks, and its own judgement, to fill the space until a more specific instrument exists.

Why Claims Desks Are the Next Logical Target

Deepfake fraud targets whatever channel pays out fastest against the weakest verification. Investment scams exploit trust in a familiar face. Insurance claims exploit trust in a photograph, a repair estimate, or a claimant's own voice on a recorded call, and insurers outside the Caribbean are already living with the consequences. Pindrop's 2024 Voice Intelligence and Security Report, drawn from more than 1.2 billion analysed calls, found that synthetic voice fraud attacks against insurance companies rose 475% that year, well ahead of the 149% increase banks recorded over the same period. Separate research from Sumsub found deepfake-related fraud across financial services, insurance included, grew roughly 3,000% between 2022 and 2025.

Set against the wider size of the problem, the incentive is obvious. The Coalition Against Insurance Fraud estimates that fraud accounts for roughly 10% of US property-casualty insurance losses, totalling more than $308.6 billion a year. Generative tools that can fabricate an injury photo, a repair invoice, or a claimant's voice do not need to succeed often to be worth running at scale, and unlike a forged paper document, a synthetic voice or video can be produced in minutes for a few dollars.

Nothing in the public record shows a Caribbean insurer has yet suffered a confirmed deepfake claims fraud loss. That absence is worth taking seriously rather than as reassurance. The same was true of Trinidad's investment-fraud scammers in July 2025, a month before the first alert went out.

The Boardroom Is Not Ready Either

Whether an insurer catches this risk early depends partly on whether its board is paying attention, and the Caribbean's own directors say most are not. A PwC Caribbean Corporate Governance Survey of 154 board directors across The Bahamas, Barbados, Grenada, Jamaica, Saint Lucia, and Trinidad and Tobago, published in March 2026, found that only 6% believe their boards spend enough time understanding AI's impact, and just 9% believe they receive sufficient information to address the risks it introduces. Those figures cover AI governance broadly, not insurance fraud specifically, but a board that is not confident in its general AI oversight is not well positioned to have set a specific deepfake-detection policy for its claims function either.

That gap sits underneath the industry's own AI adoption, not just its exposure to fraud from outside. The same generative tools an insurer might use to speed up underwriting or draft a claims summary are, in a slightly different configuration, the tools a fraudster uses against it. A board with limited AI literacy is poorly placed to ask the right questions about either use.

CAIRMC's Answer: The CARA Framework

The Caribbean AI Risk Management Council built the Caribbean AI Risk Assessment (CARA) framework to close exactly this kind of gap between a broad regional roadmap and a sector's daily operating decisions. CARA starts with a full inventory of an organisation's AI systems, then classifies both those systems and AI-enabled threats into four risk tiers, using logic aligned to the EU AI Act rather than a Caribbean-specific standard built from nothing. Applied to a claims department, it gives a compliance or investigations team a repeatable way to classify a suspected deepfake or synthetic document, instead of improvising a fresh response every time a new fraud pattern shows up.

CAIRMC is chaired by Adrian Dunkley, who also founded StarApple AI in Kingston in 2023, generally credited as the first dedicated AI company established in the Caribbean, and who serves as president of the Caribbean AI Association. Caribbean Insurance is part of the same Caribbean AI ecosystem those organisations built, which is worth disclosing plainly rather than treating as incidental. CAIRMC's own position, stated publicly, is that a framework built for the region's regulatory realities, rather than adapted after the fact from a larger market's rules, is what lets a small insurer or a single-territory regulator act on an AI risk without waiting years for a CARICOM-wide standard to catch up.

What This Means for Your Policy

Most readers of this article are policyholders, not board directors. What the last few sections mean in practice comes down to a short set of habits worth building now, while the Caribbean's specific deepfake-claims-fraud incident is still theoretical rather than reported.

1. Expect more verification steps on high-value or unusual claims, not fewer. An insurer asking for a second form of proof, or a call-back through a known number rather than the one that contacted you, is responding to a real industry trend, not being difficult.
2. Ask your insurer or broker whether they have a documented policy on AI-generated content in claims. Not every insurer will have an answer yet, but the question is fair, and it establishes a record either way.
3. Treat an unsolicited video, call, or message asking you to move money or share policy details as suspicious by default, even if the face or voice looks and sounds correct. Trinidad's six alerts all involved content convincing enough to fool people who had no reason to doubt it.
4. Keep your own dated records, photographs, and receipts independent of any system an insurer or scammer might touch. A synthetic claim is easier to contest when the genuine version of events is already documented on your side.

None of this requires treating AI in insurance as a threat rather than a tool. The same technology speeding up legitimate underwriting and claims processing across the region is the technology fraudsters are starting to repurpose. The practical difference between the two uses is verification, and verification is the one thing a regional roadmap cannot install on its own. That still has to happen inside each insurer, each claims department, and each policyholder's own habits.

Frequently Asked Questions