As digital transformation accelerates and more devices are connected through the Internet of Things (IoT), the attack surface keeps expanding, and cyber attacks are becoming more frequent and sophisticated. Firms also face growing liability for data breaches, including compensation claims and regulatory fines. At the same time, AI has become one of the most powerful tools for identifying nonlinear patterns in data and improving predictions. Recent advances such as large language models (LLMs) and computer vision models are now widely used to boost productivity and support decision-making. Yet AI is also a new source of risk. Beyond model complexity and uncertainty, agentic AI systems can act on their own, so a single error can lead directly to real losses such as wrong transactions, business interruption, or liability claims.

Demand for insurance against these risks is growing, but designing and pricing such coverage is difficult. Traditional actuarial models assume ample historical loss data and independence between risks, and cyber and AI risks satisfy neither assumption. Loss data are scarce, while threats evolve quickly. Because many firms rely on the same networks, software, cloud services, and AI models, one incident can cause losses at many firms at once (accumulation risk). Losses are also heavy tailed, so a few extreme events can dominate total losses. For AI, it is often unclear who is liable when something goes wrong: the developer, the operator, or the user.

Addressing these challenges requires new ways to measure, price, and transfer cyber and AI risks. Our research includes:

• Cyber loss prediction and heterogeneity analysis
• Measurement of the financial impact of cyber incidents
• Analysis of extreme and systemic cyber loss events
• Cyber risk transfer (cyber insurance, parametric insurance, reinsurance, catastrophe bonds)
• AI-driven risk modeling
• Modeling and insurability of AI risks

Climate change affects both the real economy and the financial system through two main channels. Physical risk refers to losses from acute hazards such as typhoons, floods, and heatwaves, and from chronic changes such as rising temperatures and sea levels. Transition risk refers to financial losses that arise as the economy shifts toward a low-carbon economy, for example through stricter carbon regulation or the spread of low-carbon technologies.

As natural disasters become more frequent and severe, insured losses are rising. The protection gap, the share of economic losses not covered by insurance, is also widening. Quantifying how climate change affects losses is therefore essential for pricing, solvency management, and supervisory stress testing.

Measuring these effects, however, is challenging. Historical loss data reflect past climate conditions and may not represent future risk. The period in which climate change effects have been clearly observed is short, and climate scenarios themselves are highly uncertain. Moreover, final losses depend not only on the hazard but also on the people, buildings, and infrastructure exposed to it and how vulnerable they are. Climate data must therefore be combined with infrastructure and socio-economic data. Transition risk is equally uncertain, as it depends on factors such as policy shifts after changes in government and the pace of carbon-reduction technology.

We therefore develop models that link climate hazards to economic and insured losses while explicitly accounting for uncertainty. Our research includes:

• Physical risk modeling (acute & chronic hazards)
• Climate stress testing for (re)insurers
• Catastrophe risk transfer (reinsurance, catastrophe bonds)
• Climate insurance and the protection gap
• Transition risk: carbon regulation and ESG

Insurance pools risks across many policyholders, so that losses too large for any single household or firm can be shared. In this way, it plays a key role in building resilience to large-scale disasters. This area studies the decisions of insurance buyers and insurers, and how insurers measure and manage risk.

Despite its importance, insurance uptake remains low in many markets. To raise it, we need to understand who buys insurance and why others do not. At the same time, insurers must measure new and evolving risks accurately in order to set fair premiums and hold adequate capital.

Conventional statistical methods, however, face several limitations:

• Heterogeneity: Policyholders and insurers differ widely in size, preferences, and financial conditions, so analyses based on averages can miss important differences.
• Extreme losses: Insurance and financial data contain rare but very large losses that standard distributional assumptions fail to capture.
• Dependence: Most models assume that risks are independent. Yet risks such as pandemics, natural catastrophes, and cyber attacks can hit many policyholders at once or spread from one to another. When such losses grow large enough to threaten the stability of the whole system, they become systemic risk. Ignoring this dependence can lead to underestimating or overestimating losses and required capital.

New methodologies are therefore needed to measure these risks more accurately and consistently. Our research includes:

• Heterogeneity in insurance and reinsurance demand
• Factor-based risk measures
• Dependence modeling and copula approaches
• Systemic risk and risk aggregation for capital requirements
• Simulation-based risk analysis
• Insurer governance, ESG, and efficiency

Life insurance covers risks related to human life (death, survival, and health), and many contracts run for decades. Because there is a long gap between when premiums are received and when benefits are paid, life insurers are highly exposed to changes in the economic environment, especially interest rates. Under IFRS 17, insurance liabilities are now valued using current market interest rates. This makes asset-liability management (ALM), the joint management of asset and liability cash flows, even more important. In addition, as populations age, annuities play a growing role in providing stable retirement income.

Long-term risks, however, are hard to predict. Life expectancy keeps rising with advances in medicine, while new factors such as climate change (e.g., heatwaves) affect mortality and morbidity. As a result, assumptions based on historical data may not hold in the future. Longevity risk, the risk that people live longer than expected, places a heavy burden on annuity providers. Because this risk interacts with interest rate risk over several decades, even small model errors can lead to large losses.

To address these challenges, our research includes:

• Mortality and longevity risk modeling
• Annuities and retirement income
• Asset-liability management (ALM) and interest rate risk
• Economic scenario generation for insurance valuation
• Health insurance claims prediction