Edition 044 • September 18, 2026

The Credibility Report

Actuarial Intelligence for Insurance Professionals

What's in this edition

MembersCap's new cat bond fund, SCOR's board-level AI warning from Monte Carlo, AM Best's resilient Latin America read, sidecar structuring lessons from Northern Re, Swiss Re's $300bn Florida scenario — and foundation models for insurance risk headlining a strong arXiv week. All deduplicated against every prior edition.

Headlines (primary sources)

MembersCap launches cat bond fund, was sole investor behind innovative cat / cyber ILW

Artemis.bm

Members Capital Management (MembersCap) has launched its first catastrophe bond fund strategy to expand its offering to clients — and disclosed it was the sole investor behind a recent innovative industry-loss warranty covering catastrophe and cyber risk. The Bermuda-regulated ILS manager launched in 2024.

Read source ->

SCOR leans into AI as reinsurer warns of governance and liability challenges

Reinsurance News

Speaking at a Monte Carlo Rendez-Vous media briefing, SCOR Group CEO Thierry Léger and SCOR Global P&C CEO Jean-Paul Conoscente set out the reinsurer's view of the forces reshaping the industry, with artificial intelligence — and its governance and liability questions — among the most significant themes.

Read source ->

Latin American reinsurance market maintains growth despite prolonged soft cycle: AM Best

Reinsurance News

AM Best's September 2026 Market Segment Report finds the Latin American reinsurance market has remained resilient despite weaker economic growth, significant catastrophe exposures and a prolonged soft market — with continued premium growth, abundant reinsurance capacity and opportunities to close the region's substantial protection gap.

Read source ->

Sidecar growth signals strong investor appetite, but structure remains key: Anthony McKelvy, Northern Re

Artemis.bm

While growth in sidecar capital reflects healthy demand for insurance-linked returns, long-term success will depend on the sophistication of the structures sitting behind carrier balance sheets, Anthony McKelvy, Managing Partner of Northern Re, told Artemis around the 2026 Monte Carlo Rendez-Vous.

Read source ->

Population growth and asset concentration push Florida hurricane risk to $300bn: Swiss Re

Reinsurance News

Swiss Re Institute research suggests a category 5 hurricane striking Miami or Tampa Bay today could generate insured losses of $300 billion or more, as population growth and rising asset concentration in exposed areas drive up potential insured natural catastrophe losses worldwide.

Read source ->

Research Spotlight (arXiv)

Towards foundation models for insurance risk modelling

arXiv • Score: 30 • Paper of the Month

Claim narratives, images and sensor data hold risk information that current actuarial models struggle to use. This paper explores how foundation models — pretrained at scale, then adapted per task — could turn those high-dimensional sources into model-ready variables, such as flagging a worsening injury in a claim note so a reserving model can react, potentially reducing the experience needed to develop each application.

Open paper ->

STUART: Sequence Triage and qUAntification of Read Transcripts for Rapid Ionizing Radiation Exposure Assessment

arXiv • Score: 29

A mapping-free machine learning framework for rapid biological dosimetry after radiation exposure, converting raw sequencing reads into k-mer profiles to bypass computationally intensive alignment — an NLP-inspired triage architecture built for mass-casualty scenarios on mobile hardware.

Open paper ->

Quantifying the 2027 Solvency II Risk Margin Reform

arXiv • Score: 28

An analytical and numerical framework for the 2027 reform that cuts the risk-margin cost-of-capital rate from 6% to 4.75% and adds time-dependent attenuation of future SCRs — normalizing discounted projected capital requirements to obtain an exact representation of the revised-to-previous risk-margin ratio, with sharp universal and horizon-specific bounds.

Open paper ->

WaVeFuse: Regime-Adaptive Equity Index Forecasting via Channel-Wise Wavelet Denoising and Vertical Attention Fusion

arXiv • Score: 28

A dual-branch deep learning architecture for equity index forecasting that pairs Symlet-4 wavelet denoising of OHLCV data with channel-wise wavelet encoding and adaptive vertical attention fusion — directly targeting noise propagation into technical indicators, indiscriminate multi-scale decomposition, and static fusion that cannot track regime shifts.

Open paper ->

Multivariate Ruin in Lévy-driven Insurance Risk Models under Heavy Tails

arXiv • Score: 27

Finite-horizon ruin analysis for a multi-line insurer with heavy-tailed claims modeled as a multivariate increasing Lévy process — covering insolvency events ranging from single-line failure to simultaneous failures of groups of lines linked through internal transfers, reinsurance, or guarantee arrangements.

Open paper ->

Null importance: Disentangling relevance for interpretable machine learning

arXiv • Score: 26

A unified population-level characterization of when a feature is irrelevant under different notions of relevance — marginal and conditional dependence, predictive risk, functional invariance, and causal effects — disentangling what "importance" actually means across interpretable ML applications.

Open paper ->

Practical Takeaways

  • For ILS and alternative capital watchers, MembersCap's cat bond fund launch and McKelvy's sidecar commentary signal a maturing offered market — structural sophistication is becoming the differentiator, not just appetite.
  • For AI adoption programs, SCOR's Monte Carlo message is that AI's governance and liability questions are now board-level topics — treat model risk management for AI as a front-office issue, not a back-office one.
  • For Solvency II capital teams, the 2027 risk-margin reform paper's exact revised-to-previous ratio bounds are directly usable in ORSA and capital planning ahead of the recalibration.
  • For pricing and reserving teams, the null-importance framework offers a defensible vocabulary for which features — and which notion of relevance — actually drive a model, beyond single-score importance rankings.

What We're Watching

  • arXiv — foundation models for insurance risk and Solvency II quantification head a strong actuarial week
  • Artemis.bm — ILS expansion into cat bond funds and the sidecar structuring debate out of Monte Carlo
  • Reinsurance News — post-Rendez-Vous signal-taking: AI governance, soft-cycle discipline, Latin American growth
  • Swiss Re Institute — accumulating nat-cat exposure concentration studies, with Florida at $300bn

Until next time, stay credible.

- The Credibility Report

Edition 044 | Prepared September 18, 2026 (UTC)