$100 Billion, 90% Silent, 0 National AI Strategies: The Caribbean's Insurance Blind Spot on AI Risk

A Report Published Nine Days Before the Region's Own Roadmap

The finding itself is not subtle. Research from the Artificial Intelligence Underwriting Company, produced jointly with researchers from Anthropic and OpenAI and reported by the Financial Times on Tuesday, 14 July 2026, found that more than 90% of insurers' exposure to artificial intelligence sits inside conventional policies that were never explicitly priced for it. The risk is, in the report's own language, largely unpriced and in many cases unnoticed by the carriers holding it. AIUC co-founder Rajiv Dattani put the commercial consequence plainly: businesses cannot adopt AI unless they know the risk has been quantified and managed, and right now most of the industry underwriting that adoption cannot make that claim.

The scale attached to that admission is what makes it an investment story rather than a compliance footnote. AIUC estimates a serious AI catastrophe could cause around $100 billion in direct damages, with the larger economic fallout, if insurers pull coverage, businesses cool on AI adoption, and investors retreat from the sector, potentially reaching into the trillions. The litigation record already exists to justify that concern: Google faces a suit seeking at least $110 million over allegedly defamatory AI Overviews output, and Air Canada remains the reference case regulators cite for a chatbot that invented a discount the airline was later ordered to honor. FINRA's 2026 compliance report flagged hallucination risk directly, telling broker-dealers to establish procedures for AI agents that may act beyond a user's intended scope. Lloyd's of London has already responded commercially, introducing an AI hallucination-specific insurance product rather than waiting for the exposure to surface as a claim.

What Silent Actually Means on a Policy Schedule

Silent AI exposure is not a hypothetical clause. It describes the gap between a policy that says nothing about AI and a court willing to read that silence in the policyholder's favor. Most professional indemnity, product liability, cyber, and D&O wordings predate the generative AI boom entirely, and their services and exclusions sections simply do not contemplate an AI system as the cause of loss. As Nicholas Blackmore of Kennedys' Melbourne office put it in an earlier analysis of the same trend, that type of scenario may simply be covered under the policy, but the insurer then has a problem if it did not price the risk in. That is the exact mechanism the AIUC report quantifies at industry scale.

Carriers are already responding, and the response is exclusion rather than expansion. In November 2025, AIG, Great American, and W.R. Berkley filed requests with US state insurance regulators to carve AI-related liability out of general liability and errors-and-omissions policies. Those filings are being reviewed jurisdiction by jurisdiction, and the direction across the market is unambiguous: carriers that can identify their silent AI exposure are moving to either exclude it outright or price a narrow, explicit AI endorsement on top. Carriers that cannot identify it, which the AIUC report suggests is most of the market, remain exposed to whatever a court decides later.

154 Directors, Six Countries, One Governance Statistic

The Caribbean version of this gap starts one step further back, at the board level. A March 2026 PwC survey of 154 directors across six Caribbean nations found that only 6% believed their boards spend sufficient time on AI oversight, and only 9% said they receive adequate information to address AI-related risk. Those are the same boards responsible for approving AI deployment inside the region's banks, insurers, and credit unions. A board that cannot get adequate information on AI risk in general is not positioned to identify, ahead of a claim, whether its own professional indemnity or D&O coverage has quietly become silent AI exposure in the AIUC sense.

The regulatory backdrop compounds it. As of April 2026, no CARICOM member state had published a standalone national AI strategy. Data protection enforcement, the baseline legal infrastructure that would normally anchor any AI governance regime, sits in three uneven tiers across the region: Jamaica and Barbados have operational authorities with functioning enforcement and, in Jamaica's case, a 72-hour breach notification requirement in force since December 2023; a second tier, including Trinidad and Tobago, Guyana, and Saint Lucia, has partial legislation; and a third tier, including Haiti, Dominica, and Saint Kitts and Nevis, has minimal or no enforceable framework at all. A carrier writing risk anywhere in that third tier is underwriting AI exposure with even less regulatory scaffolding beneath it than the global market the AIUC report describes as already unprepared.

Where the Exposure Already Sits Inside Caribbean Books

This is not a theoretical future risk for the region. Financial institutions across Jamaica, Trinidad and Tobago, and Barbados already deploy AI for fraud detection, credit scoring, and customer service automation, the same functions the AIUC report identifies as sources of silent exposure elsewhere. The Caribbean Catastrophe Risk Insurance Facility, CCRIF SPC, is instructive by contrast: its 39 members, 19 Caribbean governments, 4 Central American governments, and a set of regional electric and water utilities, are covered through a parametric structure that pays out against a rules-based trigger rather than a discretionary claims process, and CCRIF's own public materials make no mention of AI in its claims or risk modeling. CCRIF itself is not the exposure. The exposure sits one layer down, in the primary and commercial insurers writing professional indemnity, liability, and D&O policies across the region, the same book of business the AIUC report describes as globally unpriced, and those insurers draw reinsurance capacity from the same international carrier market now filing AI exclusions.

A New Regulator's Chair, an Agenda Without the Word AI

On 2 July 2026, Dana L. Munnings-Gray, Superintendent of The Bahamas' Insurance Commission, was elected president of the Caribbean Association of Insurance Regulators (CAIR) at a conference in Belize, the second time a Bahamian regulator has held the role after an earlier term running from 2014 to 2018. The reported agenda covered regulatory developments, emerging risks, consumer protection, innovation, and regional cooperation, a set of headings broad enough to house an AI risk conversation without ever naming one. The Bahamas is now set to host both the CAIR and CAPS conferences in 2027, which gives the region's insurance regulators a full annual cycle in which AI risk pricing could become an explicit agenda item rather than an implied one folded into "emerging risks."

Eight Days to CAITF, and a Roadmap Written Before the Insurance Question

The CTU Caribbean AI Task Force, launched 18 July 2025, will formally publish its Final Report at the inaugural Caribbean AI Forum 2026, running 23-24 July at the University Inn Conference Centre, University of the West Indies, St. Augustine, Trinidad and Tobago, under the theme "AI for Caribbean Transformation: Governance, Innovation and Resilience for a Shared Digital Future." The interim report's five priority areas, regional AI governance, data sovereignty and digital infrastructure, innovation and industry development, human capacity and AI literacy, and sustained multi-stakeholder engagement, are the right categories for a first regional framework. None of them names insurance, actuarial pricing, or financial-services AI risk specifically. That is not a criticism of the drafting; a first harmonisation document has to set a governance floor before it can address sector-specific pricing questions. It does mean the region's flagship AI governance document, the one every CARICOM regulator will reference for the next several years, was substantially written before the industry it eventually has to govern admitted, in the same month, that it cannot price its own AI risk.

Risk Factor	Severity	What It Means Here
Unpriced silent AI exposure in existing policy books	High	Caribbean carriers writing PI, liability, and D&O business inherit the same unpriced exposure the AIUC report quantifies globally, with less actuarial capacity to model it locally
No national AI strategy in any CARICOM state	High	Regulators have no baseline framework to require AI risk disclosure, audit trails, or reserve adequacy from insurers or the institutions they cover
Board oversight gap (PwC, March 2026)	High	Only 6% of surveyed directors are satisfied with board AI oversight and only 9% get adequate risk information, weakening the first line of defense against silent exposure
CAIR's AI agenda lag behind global carriers	Medium	The region's insurance regulator body has not yet named AI explicitly on a public agenda, even as AIG, Great American, and W.R. Berkley reshape US policy wording
CAITF roadmap scope gap on financial services	Medium	The Final Report's five priority areas do not name insurance or actuarial risk pricing specifically; sector guidance will likely need a follow-on process
Uneven data protection enforcement across CARICOM	Medium	A three-tier legal landscape means AI governance readiness varies sharply by jurisdiction, complicating any single regional insurance standard

The Investable Layer: Quantification Infrastructure, Not Prohibition

Betting on which global carrier excludes AI risk fastest is not a Caribbean-specific trade. The more durable opportunity for regionally focused capital is in the tooling that lets an insurer, a regulator, or a reinsurer actually see and price the exposure instead of writing around it. Maestro AI Labs' own Harmonics agent framework, covered in detail in our companion piece on AI agents for regulated entities, was built for exactly this problem: agents that log every decision, escalate rather than guess, and leave a human override and an audit trail behind them, which is the underlying infrastructure an underwriter needs before it can price an AI system rather than exclude it. Global Safety Score's portable risk data layer and Credit Garden's demonstration that imported global models underperform badly on Caribbean-specific data are the same pattern applied to adjacent products: the region's AI risk, whether in credit, safety, or insurance, is systematically mispriced by models built for someone else's dataset.

Three specific structures follow from that. First, AI risk quantification and audit-trail vendors built for the compliance standard a Caribbean regulator will eventually require, not retrofitted from a US or UK product. Second, actuarial and underwriting models trained on regional loss and claims data rather than imported tables that, as the Caribbean's credit-scoring experience already shows, fail badly outside the markets they were built on. Third, a parametric-style pooled structure modeled on CCRIF's own architecture, extended from weather perils to AI-related liability, which would let smaller Caribbean insurers access AI risk capacity without each one separately negotiating exclusions with a reinsurer. None of these require waiting for a hyperscaler-scale deal to close; each is buildable now, against a regulatory gap that both the AIUC report and CAITF's own roadmap confirm is real.

Why the Sequencing Cuts Both Ways

The two dates that frame this piece, 14 July and 23 July, are eight days apart, and the order matters more than either headline alone. Global insurers have just admitted, in public, that a decade of AI deployment outran their own pricing models. The Caribbean's regulators have not made an equivalent public admission, largely because the region's task force has not yet reached the point in its mandate where insurance-specific guidance is due. That sequencing is an opening rather than a delay. A regulator, an insurer, or an investor building AI risk infrastructure for the Caribbean today is building it before a comparable exclusion wave forces the issue here the way it already has in the United States, and before a claim, rather than a report, sets the terms of the conversation.

StarApple AI, founded by Adrian Dunkley in 2023 as the first AI company established in the Caribbean and recognised across the region as the work of the Caribbean's leading AI authority, built Maestro AI Labs as its investment intelligence arm specifically to track sequencing gaps of this kind, where a global admission and a regional roadmap land in the same month without ever citing each other. This analysis draws on that ongoing tracking work, cross-checked against the report record cited throughout, rather than on either institution's press materials alone.

Related Reading Across the Caribbean AI Network

This analysis draws on and connects to ongoing coverage across the wider Caribbean AI network. Worth reading alongside it: