Trinidad Faked Four of Its Own Leaders. The Caribbean's AI Roadmap Never Says the Word Deepfake.

Trinidad and Tobago's securities regulator has now issued six separate public warnings about the same crime. Fraudsters keep generating video of the country's own leaders, past and present, endorsing investment schemes those leaders never touched. A sitting Finance Minister. A former president. A former prime minister. A former bank chairman. Even a journalist's identity, cloned to make the scam look like real reporting. Three weeks before the sixth alert went out, CARICOM's trade ministers sat down in a special session and formally adopted a regional roadmap for artificial intelligence. That roadmap runs to dozens of pages on governance, culture, and education. It does not once use the word deepfake.

Six Alerts, Four Faces, One Pattern

The Trinidad and Tobago Securities and Exchange Commission's warnings tell a specific, checkable story, not a vague one. Investigators first flagged AI-generated content impersonating the country's Finance Minister in August 2025. By October, the Senate had identified fabricated video targeting former President Anthony Carmona and former Republic Bank chairman Dr Ronald Ramkissoon. The pattern widened through 2025 and into 2026, and on 26 June the regulator issued its sixth alert, this time naming fabricated content built around former Prime Minister Dr Keith Rowley. Along the way, scammers also cloned the identity of Guardian Media editor Kejan Haynes, using a real journalist's face and voice to make a fake investment pitch look like a real news segment.

Every one of these videos exists to do one thing: put a trusted Trinidadian face in front of a fraudulent investment offer. That is not a hypothetical AI risk the region needs to prepare for. It is a live crime pattern that has run for more than a year, against a former head of state, a former head of government, and a serving cabinet minister, in a country of 1.4 million people.

Trinidad and Tobago did not sit still. In September 2025, Cabinet approved an Inter-Ministerial Steering Committee on Cyber Security and AI, and the country stood up the Caribbean's first cabinet-level AI ministry, the Ministry of Public Administration and Artificial Intelligence, under Minister Dominic Smith. That is a real institutional response, faster than most governments manage on a genuinely new threat. It is also, six alerts in, not yet a solved problem.

A Regional Roadmap That Skips the Crime Actually Happening

On 7 July 2026, CARICOM's Council for Trade and Economic Development, in its ICT configuration, formally endorsed the UNESCO Caribbean Artificial Intelligence Policy Roadmap. It is a genuinely substantial document, built from consultations with more than 1,000 institutions and individuals across 20 English- and Dutch-speaking Caribbean territories, and organised around four pillars: culture and creativity, governance and transformation, education and upskilling, and resiliency and sustainability. That is real regional coordination work, and it matters that CARICOM member states signed onto a shared approach rather than 20 separate ones.

Read the governance pillar closely, though, and the gap becomes obvious. It calls for "a regulatory framework, outlining roles and responsibilities," and it addresses fairness and accountability in AI systems built by legitimate institutions: banks, governments, telecoms. What it does not do is name deepfake investment fraud as a category of harm, or set out the verification protocols and cross-border enforcement cooperation that catching a scam video impersonating a former president actually requires. The roadmap governs how the region's own AI systems should behave. It has far less to say about criminals weaponising AI against the region's own citizens, which is the exact crime Trinidad's regulator has now documented six times.

That is not a reason to dismiss the roadmap. A shared regional governance document is worth having, and getting 20 territories to agree on four pillars is not a small achievement. But a document that took two years to build, drawing on a thousand consultations, and still does not name the fraud pattern already running against three former heads of state and a sitting minister, tells you something about the gap between policy timelines and criminal ones. Fraudsters do not wait for a roadmap to be endorsed. They were already six alerts deep before the ink dried.

The Caribbean Broadcasting Union Got Closer to the Real Problem

Twelve days after CARICOM's endorsement, a more useful diagnosis came from an unexpected room. At the Caribbean Broadcasting Union's 57th Annual General Assembly, held on 19 August 2026 in Liliendaal, on Guyana's East Coast Demerara, Guyana's Minister Kwame McCoy told the region's broadcasters something the roadmap had not quite said out loud: the Caribbean does not yet own the data that will determine how AI represents Caribbean people.

McCoy's keynote framed the problem in terms of "sovereignty, data ownership, algorithmic control, platform accountability, and the economic value of Caribbean content." He warned, correctly, that synthetic voices, images, and video make it harder by the month to tell authentic information from fabricated content, and that a fake headline or a cloned voice can shift public opinion long before any factual correction catches up. His specific recommendations were concrete rather than aspirational: train journalists in data journalism, digital verification, metadata analysis, geolocation, and synthetic media detection; have the CBU set clear standards on AI disclosure and newsroom accountability; and, most importantly, build regional verification mechanisms and datasets that ensure Caribbean people, cultures, and experiences are represented accurately in the AI systems now being trained.

Put McCoy's speech next to the TTSEC alerts and next to the CARICOM roadmap, and a clearer picture forms than any one of the three gives on its own. The region has the institutional will to respond to AI harm. What it has not yet built is the data infrastructure that response actually depends on: verified reference footage and voice samples of public figures, a shared registry of confirmed fabrication cases, and detection models trained on Caribbean faces, accents, and institutions rather than retrofitted from datasets built somewhere else.

Why This Is a Data Problem Before It Is a Policy Problem

I built StarApple AI in 2023 on a premise that this moment now proves out in the worst possible way: a global AI system trained on global data will always underperform on Caribbean-specific problems, because it was never built to recognise Caribbean-specific patterns in the first place. That premise is usually argued through insurance pricing or credit scoring. Deepfake detection is the same argument with higher stakes.

A synthetic-media detector trained mostly on North American and European faces, accents, and video formats does not automatically get better at flagging a fabricated clip of a Trinidadian former president speaking Trinidadian English on a low-bandwidth upload. Detection models learn from the examples they are shown. If the training data barely includes Caribbean public figures, Caribbean broadcast formats, and Caribbean speech patterns, the model's confidence on exactly the videos TTSEC has been flagging will be lower than its confidence on a fabricated clip of a US senator. That is not a defect in the technology. It is what happens when a region imports detection tools instead of building the reference dataset those tools need to work well on its own citizens.

This is the deeper version of the numbers worth sitting with. Deepfake fraud now makes up 6.5 percent of all fraud attempts globally, up from 0.1 percent in 2022, according to identity verification firm Sumsub. Roughly 8 million deepfakes are estimated to be circulating online in 2026, sixteen times the 500,000 estimated in 2023, according to research firm DeepStrike. Documented global losses from deepfake fraud reached at least $1.28 billion in 2025, and researchers who track the figure describe it explicitly as a floor, since more than 80 percent of reported incidents never disclose a dollar amount at all. Every one of those trend lines points the same direction: up, fast, and mostly undercounted. A region that starts building its own verification infrastructure now is building against a curve that gets steeper every quarter it waits.

The Caribbean AI Risk Management Council, which I helped establish specifically to build risk frameworks calibrated to Caribbean regulatory and economic conditions, has already published its own analysis of the gap between the CARICOM roadmap and Trinidad's deepfake wave. Work like that, done by a Caribbean institution reading Caribbean incidents, is exactly the raw material a regional verification dataset would be built from. The Caribbean AI Association exists to connect the practitioners who would build it across territories rather than leaving each country to solve the problem alone. Neither organisation can build that infrastructure from a conference room. It gets built the way StarApple AI's own platforms got built: sector by sector, dataset by dataset, starting from a real incident and a real institution willing to own the work.

What Caribbean Institutions Should Do Before the Next Alert

Waiting for a follow-up roadmap that names deepfake fraud explicitly is not a strategy, because the crime is not waiting either. There are concrete steps a bank, a ministry, a media house, or a company can take this quarter, with no regional agreement required first.

Any organisation whose executives or board members could plausibly appear in a fabricated endorsement video needs a written verification protocol now: a defined process for confirming that a video, voice note, or image purporting to show a named leader authorising a transaction is real before anyone acts on it, and a public, standing notice of the official channels the organisation will and will not use to solicit money or share instructions. Communications and compliance staff need training in reverse-image search, metadata inspection, and basic synthetic media red flags, the same competencies McCoy called for in journalism, because the exposure is identical whether the fabricated video targets a newsroom or a bank's customers.

Regional coordination has to move past a shared vocabulary and into a shared dataset. That means TTSEC's confirmed fabrication cases, and equivalent incidents from financial regulators and media houses across other CARICOM states, feeding a common, structured registry rather than sitting in six separate press releases. It means Caribbean AI companies, universities, and risk bodies contributing verified reference material rather than each rebuilding the same detection baseline from scratch. And it means the next generation of Caribbean AI builders, the students coming through programmes like The Genius Project, learning synthetic media detection as a core skill rather than a specialty, because by the time they are in the workforce the 8 million deepfakes online today will look like a modest starting number.

None of this requires CARICOM to reopen a two-year policy process. It requires Caribbean institutions to treat data ownership the way Trinidad treated the institutional response: fast, specific, and built before the next alert rather than after it.

Frequently Asked Questions

What happened with deepfakes impersonating Trinidad's leaders?

Since mid-2025, fraudsters have circulated AI-generated videos impersonating Trinidad and Tobago's sitting Finance Minister, former President Anthony Carmona, former Republic Bank chairman Dr Ronald Ramkissoon, and former Prime Minister Dr Keith Rowley, all promoting fake investment schemes. The Trinidad and Tobago Securities and Exchange Commission has issued six public investor alerts about the pattern, most recently on 26 June 2026. A Guardian Media editor's identity was also cloned to lend the scams false credibility.

Does the CARICOM AI Policy Roadmap address deepfake fraud?

Not by name. CARICOM's COTED-ICT formally endorsed the UNESCO Caribbean AI Policy Roadmap on 7 July 2026, built on four pillars covering culture and creativity, governance, education, and resiliency. The Governance pillar calls for a regulatory framework outlining roles and responsibilities, but the document addresses fairness and accountability in AI systems built by legitimate institutions. It does not specifically name deepfake investment fraud or set out verification and cross-border enforcement protocols for AI-generated impersonation crimes.

How common is deepfake fraud globally in 2026?

Deepfake fraud now accounts for 6.5 percent of all fraud attempts globally in 2026, up from 0.1 percent in 2022, according to identity verification firm Sumsub. Roughly 8 million deepfakes are estimated to circulate online in 2026, up from about 500,000 in 2023, according to research firm DeepStrike. Documented global deepfake fraud losses reached at least $1.28 billion in 2025, a figure researchers describe as a floor rather than a full total, since more than 80 percent of reported incidents carry no disclosed dollar amount.

What did the Caribbean Broadcasting Union say about AI and journalism?

At the CBU's 57th Annual General Assembly in Guyana on 19 August 2026, Guyana's Minister Kwame McCoy urged Caribbean media houses to shape AI standards rather than import them, warning that synthetic voices, images, and video can shift public opinion before a factual correction ever reaches an audience. He called for journalist training in digital verification, metadata analysis, and synthetic media detection, and for the region to build its own datasets so Caribbean people and culture are represented accurately in the AI systems being trained right now.

How is Trinidad and Tobago responding to AI-driven fraud at the government level?

Trinidad and Tobago created the Caribbean's first cabinet-level AI ministry, the Ministry of Public Administration and Artificial Intelligence, under Minister Dominic Smith. On 11 September 2025, Cabinet approved an Inter-Ministerial Steering Committee on Cyber Security and AI to coordinate national policy and report on threats quarterly. The securities regulator has also issued repeated public alerts naming the specific fabricated content as it appears, which gives citizens a direct, checkable reference point rather than a general warning.

What should Caribbean businesses and media houses do about deepfake risk now?

Wait for a regional framework and the exposure keeps compounding. Practical steps available now include building an internal verification protocol for any video or audio purporting to show an executive, minister, or public figure authorising a transaction; training frontline and communications staff to run reverse-image and metadata checks; publishing a standing public notice of the official channels a company will and will not use to solicit investment; and contributing incident data to regional efforts, such as the Caribbean AI Risk Management Council, that are building the verification datasets no single company can build alone.