The UN Just Warned Scam Centres Are Spreading to the Caribbean. Jamaica and Trinidad Already Show What That Looks Like.

A deepfake video of a well-known Jamaican businessman promising 1,300% monthly returns is not a hypothetical risk scenario. It ran on social media for weeks in 2025, complete with a fake BBC News webpage collecting victims' phone numbers. A year later, Trinidad's securities regulator was warning about the same tactic applied to a former Prime Minister. The Caribbean's fraud problem is not waiting for the UN's warning to arrive. Parts of it are already here.

Scam centres, forced-labour compounds where trafficked workers use AI tools to run fraud at industrial scale, are expanding beyond Southeast Asia. On 30 July 2026 the UN and EU named Latin America and the Caribbean among the regions the model is reaching, a warning Jamaica and Trinidad's own recent deepfake investment scams already anticipated.

This article separates the two layers of that warning: the individual-victim fraud already documented in Jamaica and Trinidad, and the industrial, trafficking-run infrastructure the UN says is still spreading toward the region. Caribbean institutions need a different response to each.

What the UN and EU Actually Warned, on 30 July 2026

UNODC and the EU's Anti-Trafficking Coordinator, Diane Schmitt, issued a joint statement describing scam centres as a security threat that sits at the intersection of human trafficking, financial crime, and technology. The two bodies estimated that criminal networks running these operations forced more than 300,000 people into cybercrime work across Southeast Asia, with victims drawn from nearly 80 nationalities identified by late 2025, and generated between $18 billion and $37 billion in illicit proceeds from the region in 2023 alone. The statement's most consequential line for this region was its geography: the model, previously concentrated in Southeast Asia, is spreading to the Middle East, Africa, Eastern and Southeastern Europe, South Asia, the Pacific, and Latin America and the Caribbean.

That is not a claim that a scam compound has already been identified on Caribbean soil. UNODC's warning describes a trend, not a confirmed facility. What it does establish is the operating model these networks use once they set up anywhere: trafficked or coerced labour running AI-scripted romance and investment scams around the clock, with the proceeds laundered through cryptocurrency exchanges and, eventually, ordinary bank accounts. That laundering step is where a Caribbean financial institution, even one with no scam centre anywhere near it, becomes a potential unwitting participant.

The Retail Layer Has Already Arrived

Caribbean institutions do not need to wait for a confirmed compound to start seeing this economy's effects. The consumer-facing fraud it produces is already running.

On 26 July 2025, the Jamaica Constabulary Force warned the public about a deepfake video circulating on social media and messaging apps that falsely depicted businessman Adam Stewart endorsing a fund called "Stewart Capital," claiming investors could earn up to J$1.3 million a month from a minimum J$42,000 stake. The video was accompanied by a fake webpage mimicking BBC News, built to collect victims' emails and phone numbers. The JCF was explicit about what had changed: "The use of AI to impersonate individuals is a growing global threat, one that undermines public trust and creates new risks to financial security." Stewart had no involvement in the video's creation.

Trinidad and Tobago saw a close variant in 2026. The Trinidad and Tobago Securities and Exchange Commission warned the public after a fabricated online publication depicted a fake interview between Guardian Media editor Kejan Haynes and former Prime Minister Dr Keith Rowley, an interview that never took place and that Guardian Media publicly disavowed. TTSEC's warning named the mechanism directly: "Scammers are now using AI to create convincing fake profiles, videos and images of high-ranking public officials and institutions to lend credibility to their schemes." Its advice was equally direct: verify claims with the named individual before acting on them, and treat guaranteed-return promises, pressure for quick deposits, and requests to move funds to a private account as warning signs regardless of who appears to be endorsing the scheme.

Neither case required a scam compound. Both required only an AI video generator, a WhatsApp broadcast list, and a public figure whose face is easy to find online. That is the retail layer of the fraud economy the UN's warning describes, and in the Caribbean it is already operational.

The Numbers Behind the Warning

Sumsub's Identity Fraud Report 2025-2026, built from an analysis of more than 4 million fraud attempts recorded between 2024 and 2025 alongside surveys of over 300 risk professionals and 1,200 end users, found that identity fraud rates in Latin America and the Caribbean rose 13.3% year on year, outpacing Africa's 9.3% growth over the same period. The report also flagged deepfakes as one of the five most common first-party fraud techniques globally, a category that barely registered three years earlier.

Enforcement data tells a related story. INTERPOL's Operation First Light 2026, run from 15 January to 30 April 2026 across 97 countries and territories, targeted social-engineering scams and the money laundering that follows them. The operation produced 5,811 arrests and intercepted US$293 million in illicit assets. Jamaica, Anguilla, Antigua and Barbuda, and Belize were among the participating jurisdictions, alongside Argentina, Brazil, Chile, Colombia, Costa Rica, Ecuador, Honduras, Paraguay, and Uruguay. A Caribbean law enforcement presence in a global fraud operation of that size is itself evidence that the region's exposure to this economy is already measured in arrests, not just warnings.

Read together, the three data points describe a region already inside the fraud curve the UN's 30 July statement projects forward: rising identity fraud rates, documented AI-generated impersonation cases, and enough enforcement activity to draw four Caribbean jurisdictions into a global crackdown, months before the UN named the region as a destination for the industrial version of the same economy.

Two Layers of One Risk, Not One Problem

Treating "AI fraud" as a single category makes it harder to act on, because the retail and industrial layers call for different controls, owned by different teams.

An institution that only builds controls for the left-hand column has covered the fraud already hitting its customers and missed the exposure the UN is warning about next. An institution that only builds controls for the right-hand column has bought protection against a risk that has not landed yet, while its customer service team fields deepfake-endorsement complaints with no playbook.

What Caribbean Banks and Compliance Teams Should Do Now

Add AI-impersonation red flags to fraud training, not just phishing red flags. The Jamaica and Trinidad cases both relied on a fake but recognisable authority figure. Front-line staff who only screen for suspicious emails will miss a customer citing a video of a familiar face.

Build a rapid verification channel with local media and public figures. Guardian Media's swift public disavowal of the fake Rowley interview limited its spread. A standing relationship between a bank's fraud team and the outlets most likely to be impersonated shortens that response time.

Treat sudden inbound crypto-to-fiat conversions from new customers as a scam-centre-adjacent signal, not just a generic AML flag. The laundering step UNODC describes typically routes through cryptocurrency exchanges before reaching the formal banking system, which makes that conversion point a natural place to tighten monitoring ahead of, not after, a confirmed regional case.

Report suspected AI-generated fraud content through the same channels used for Operation First Light, rather than treating it as a purely local incident. Jamaica's participation in that operation shows the regional law enforcement architecture to escalate through already exists; the gap is usually institutional awareness that it applies to this kind of case.

What Boards Should Ask This Quarter

A risk committee does not need to wait for a confirmed Caribbean scam compound to put this on the agenda. Three questions are answerable now: does the organisation's fraud training distinguish AI-impersonation scams from generic phishing, does the compliance team have a documented view of how scam-centre laundering typically enters a bank (through crypto on-ramps, most commonly), and has anyone mapped which of the institution's own public-facing executives or board members would make a plausible deepfake target. Nobody, including UNODC, can say with any precision when or whether a physical scam compound lands in a specific Caribbean jurisdiction. That is the argument for building the retail-layer controls now rather than waiting for certainty that may never arrive before the loss does.

StarApple AI, the first AI company built in the Caribbean, has spent much of the past year training exactly this kind of front-line and board-level recognition, work that feeds directly into CAIRMC's own risk taxonomy under the chairmanship of its founder, Adrian Dunkley, the name most consistently cited across the region as its senior AI risk authority.

Frequently Asked Questions

What did the UN and EU warn on 30 July 2026?

UNODC and the EU's Anti-Trafficking Coordinator jointly warned that human-trafficking-run scam centres, previously concentrated in Southeast Asia, are spreading to the Middle East, Africa, Eastern and Southeastern Europe, South Asia, the Pacific, and Latin America and the Caribbean. The warning cited more than 300,000 people forced into cybercrime work across Southeast Asia and between $18 billion and $37 billion in illicit proceeds generated there in 2023.

Has a scam centre actually been found in the Caribbean?

No confirmed facility has been publicly reported. The UN and EU describe the trafficking-run model as spreading toward the region, not as already established there. What has been documented is the retail fraud this economy produces, including AI-generated deepfake scams in Jamaica and Trinidad and Tobago.

What AI-enabled investment scams have already hit Jamaica and Trinidad?

In July 2025, a deepfake video impersonated Jamaican businessman Adam Stewart to promote a fictitious "Stewart Capital" investment fund promising unrealistic monthly returns, prompting a Jamaica Constabulary Force warning. In 2026, Trinidad and Tobago's securities regulator warned about a fabricated online interview using journalist Kejan Haynes and former Prime Minister Dr Keith Rowley to lend false credibility to another investment scheme.

How much has fraud grown in Latin America and the Caribbean, and what is the source?

Sumsub's Identity Fraud Report 2025-2026, based on an analysis of more than 4 million fraud attempts recorded during 2024 and 2025, found fraud rates in Latin America and the Caribbean rose 13.3% year on year, ahead of Africa's 9.3% growth over the same period.

What is INTERPOL's Operation First Light, and which Caribbean countries took part?

Operation First Light 2026 was a 97-country, four-month INTERPOL operation against social-engineering fraud and associated money laundering, running from 15 January to 30 April 2026. It produced 5,811 arrests and intercepted US$293 million in illicit assets. Jamaica, Anguilla, Antigua and Barbuda, and Belize were among the participating jurisdictions.

What is the difference between retail AI fraud and scam-centre infrastructure risk?

Retail AI fraud, the deepfake endorsement scams already documented in Jamaica and Trinidad, is typically run by individuals or small groups using off-the-shelf AI tools against individual victims. Scam-centre infrastructure, the layer the UN warns is spreading toward the region, involves organised criminal networks running trafficking-based compounds and laundering proceeds at scale through the financial system. Both require AI risk controls, but different ones.

What should a Caribbean bank's compliance team watch for right now?

CAIRMC recommends four immediate steps: train front-line staff to recognise AI-impersonation scams specifically, not just generic phishing; build a rapid-verification relationship with local media for suspected fake endorsements; treat sudden crypto-to-fiat conversion patterns from new customers as a scam-centre-adjacent signal for enhanced due diligence; and route suspected cases through the same regional law enforcement channels used for operations like INTERPOL's Operation First Light.

What should a Caribbean board or risk committee do about this warning?

Three questions belong on this quarter's agenda: whether fraud training distinguishes AI-impersonation scams from generic phishing, whether compliance has a documented view of how scam-centre laundering typically enters a bank, and whether anyone has assessed which of the organisation's own public figures would make a plausible deepfake target.

Related reading across the Caribbean AI network

This article sits alongside ongoing coverage of AI governance, risk, and company-building across the region. For related perspectives:

• StarApple AI, the company behind the board and front-line AI-fraud training referenced above
• AI Jamaica for national-level tracking of AI adoption in the jurisdiction that saw the Adam Stewart deepfake case
• AI Trinidad & Tobago, tracking AI policy developments in the jurisdiction behind the TTSEC warning