Karen Clark & Co says AI could cut catastrophe model updates to days

Artificial intelligence, Karen Clark & Co argues in a new white paper, could compress catastrophe model updates from months to days. The Boston modeler calls the technology the third wave of catastrophe modeling, after statistical and physical methods; it would feed atmospheric and claims data continuously into the firm's existing physical models. Royal Gazette Bermuda Re reports on the paper.

KCC's strongest case lies in secondary perils — severe convective storms, winter storms and wildfires — where statistical models built from a limited set of variables have struggled. Severe convective storms alone produced $61 billion in insured losses globally in 2025, the third-highest total on record for the peril. KCC's current severe-convective-storm model pulls in more than 30 gigabytes of satellite, radar and weather data a day. It delivers hail, tornado and wind estimates by 7 a.m. Eastern.

Behind that model sit more than eight years of event estimates. It also draws on more than 100 terabytes of high-resolution atmospheric data. The claims data behind it runs to tens of billions of dollars. KCC says that is the training material that could teach AI to recognize complex storms, derechos included. The paper arrives as the Bermuda Monetary Authority consults on responsible AI guidance for financial services, with catastrophe modeling among the activities named. Royal Gazette also cites an analysis putting Bermuda insurers' major-disaster exposure at nearly $220 billion. That is 7.5% higher than the prior year.

For ILS investors, the stake is the pace of repricing. Cat bond spreads and collateralized reinsurance terms track model output, and that output currently moves on a months-long cycle. KCC calls the days-long refresh an expectation rather than a live capability; if the timeline holds, a storm that crosses modeled loss thresholds could show up in pricing within days. Cat bond managers have a concrete reason, then, to watch what comes out of the BMA consultation.