This article responds to the current cycle of emerging-risk survey work in the actuarial profession - a large practitioner and executive survey whose headline findings deserve to be read not as a mood board but as a set of instructions to modelling teams.
The 2026 emerging risk survey results for insurers, drawing on hundreds of risk practitioners and a substantial C-suite cohort, carry four headline signals: financial volatility and geoeconomic shifts lead the near-term worries; adverse outcomes from AI dominate the long-horizon list; climate has been reclassified from emerging to embedded; and - the finding practitioners flagged most strongly - risks are increasingly interconnected, arriving in combinations rather than singly.
Survey write-ups usually stop at the ranking. The more useful exercise is translation: if these are the risks the profession itself says are coming, what precisely should change in the models? Read that way, the survey is a modelling brief, and its work packages look like this.
Interconnection is a dependency-assumption instruction
"Risks are interconnected" is the survey's most repeated theme and its most concrete one. In model terms it says: the dependency structures carrying your aggregation - correlation matrices, copulas, scenario co-movements - are the assumptions most likely to be wrong, and wrong in the dangerous direction. Diversification benefit booked between risk classes is exactly what interconnection erodes.
The work package: inventory where diversification credit is material; challenge Gaussian tail behaviour in those places; introduce shared-driver thinking into scenario design, so that scenarios move everything a driver touches rather than one class at a time. My article on transition risk tail dependency sets out the machinery for one family of drivers; the survey's point is that the same logic applies across the board.
"Adverse AI outcomes" is a two-sided instruction
The long-horizon AI concern cuts in two directions simultaneously. As an operational and model risk: firms are deploying AI into pricing, claims, underwriting and servicing faster than validation frameworks are extending to cover it - the gap my pieces on AI model risk governance and agentic AI risk management address. As an underwriting exposure: insurers are writing the liability policies - cyber, tech E&O, professional lines, D&O - onto which the economy's AI failures will land. The survey implies both a governance work package (extend model risk frameworks now) and a portfolio one (map where AI-driven loss could enter the book, and what current wordings actually say about it). Most firms have started the first; few have seriously attempted the second.
Geoeconomic shifts are a scenario-design instruction
Geopolitical risk resists probabilistic treatment - there is no credible frequency distribution over trade ruptures. The survey's elevation of geoeconomic risk is therefore an instruction about scenario design: build fewer, deeper scenarios around named transmission channels - trade, sanctions, energy, supply chains, cyber spillover - and test them for simultaneity, since geoeconomic events are precisely the shared-driver events that hit multiple channels at once. I set out a fuller design method in my article on geopolitical scenario analysis; the survey supplies the mandate.
Financial volatility is a calibration-humility instruction
Near-term concern about financial volatility, in a world where rate regimes have already whipsawed once this decade, translates into specific model hygiene: recalibration windows that do not silently average across regimes; explicit regime-conditional parameters where behaviour differs (lapse, prepayment, credit migration); and liquidity stress assumptions revisited against the current, faster-moving deposit and surrender environment. None of this is novel methodology - it is the discipline of asking which calibrations quietly assume the previous regime.
Climate's reclassification is a scope instruction
Climate moving from "emerging" to "embedded" is the survey formally retiring the excuse for treating climate as a side exercise. The modelling consequence - climate inside routine validation scope, with the burden of proof on omission - is the subject of my article on climate risk model validation standards. In brief: materiality triage across the model inventory, non-stationarity testing on climate-sensitive calibrations, and governed treatment of vendor climate inputs.
Reading surveys as briefs
There is a general habit worth building here. Emerging-risk surveys are usually consumed as horizon-scanning colour - interesting, unactionable. But a survey of several hundred risk professionals converging on named concerns is close to the best expert elicitation the industry produces about where models will fail next. The respondents are describing, in aggregate, the weaknesses of the frameworks they themselves operate.
Translated into a single page for a modelling committee, the 2026 message is: harden dependency assumptions, extend validation to AI, design shared-driver geoeconomic scenarios, make calibrations regime-aware, and pull climate fully inside scope. Each is assignable, schedulable and testable - which is what distinguishes a brief from a mood.
Key Takeaways
- The 2026 emerging-risk survey's findings translate directly into modelling work packages rather than horizon-scanning colour.
- Interconnection is an instruction to challenge dependency structures and diversification credit, and to design scenarios around shared drivers.
- The AI concern is two-sided: extend model governance to AI systems, and map AI-driven loss potential inside underwriting portfolios and wordings.
- Geoeconomic and volatility concerns call for named-channel scenario design and regime-aware recalibration respectively.
- Climate's reclassification from emerging to embedded moves it into routine validation scope with the burden of proof on omission.
Frequently Asked Questions
What are the top emerging risks for insurers in 2026? Current professional survey work points to financial volatility and geoeconomic shifts as leading near-term concerns, adverse outcomes from AI as the dominant long-horizon risk, climate reclassified as an embedded rather than emerging risk, and - across all of these - increasing interconnection between risks.
How should insurers act on emerging risk surveys? By translating each finding into a modelling work package: dependency-structure and diversification-credit review for interconnection, validation-framework extension and portfolio exposure mapping for AI, shared-driver scenario design for geoeconomic risk, regime-aware recalibration for volatility, and validation-scope inclusion for climate.
Why does risk interconnection matter for insurance models? Because aggregation frameworks book diversification benefit through dependency assumptions calibrated to calmer periods. Interconnected risks arrive together, eroding that benefit exactly when capital matters - so the dependency structure, particularly in the tail, is the assumption most exposed to the survey's central finding.