Delegates to an insurance linked securities conference were told that critical tools used by insurers and reinsurers for catastrophe modelling can be out of date if they are more than two years old.

That position was taken by the pioneer of catastrophe models, who said that the very systems that remain the bedrock of predictive risk financing — the partial basis used by insurers and reinsurers to price catastrophe risk — is outdated, and therefore flawed.

Karen Clark, founder and CEO of Karen Clark & Company, was speaking at the ILS Bermuda Convergence 2026 conference during a panel discussion “Modelling Reimagined: Inside New Technologies and Multiple Views of Risk” at the Hamilton Princess & Beach Club.

Ms Clark said: “If your models are even two, three years old, particularly for the atmospheric peril models, you’re already behind the trends.”

The panel was moderated by Jo Syroka, director of new markets, Fermat Capital Management, and included Robert Silva, risk analytics expert and lead actuary at Zesty AI and John Wardman, the chief commercial officer at MaxInfo.

Ms Clark said that for US perils the biggest driver for increasing losses has been exposure growth — home building costs have nearly doubled in a decade and people are also building larger homes.

She said: “Climate change is impacting hurricanes, wildfires. We know how to incorporate that in the models. But still, the impact so far from climate change has been relatively small relative to the exposure growth.”

The reason, she said, people believe secondary perils, particularly severe convective storms (SCS), are rising much faster is because the models were wrong. They have been updated, and are still wrong.

“They were drastically underestimating the risk,” she said. “It's not that the risk has changed or the losses are going up, but the models are trying to catch up. The first generation models are trying to catch up. So every time they're updated, there's a big increase in losses.

“So that’s really why it’s perceived that SCS is increasing faster. But really all of the perils are increasing generally similarly because of the exposure growth — maybe hurricanes a little faster, with people still moving towards the coast.

“You do see demographics driving maybe hurricane risk up a little bit more. But even in the Midwest and other areas of the US, you do see obviously more construction, and it's really the construction cost.

“Construction cost, as a rule of thumb, generally rises at about double the CPI [consumer price index]. So double the rate of inflation. And the Covid-induced supply shortages really haven’t gone away.”

Mr Silva agreed that there were several reasons for rising losses in recent years, and referred to human intervention.

He said: “[In] the United States, greater than 85 per cent of all wildfire ignitions are caused by humans. If you go to Europe, that figure is even higher. Floods? Humans love to pave over the Earth's surface in an impermeable surface, thereby raising the probability of surface flooding.

“So it’s very interesting to me that with the influx of increasing exposure, you also have that control on hazard, and so you therefore have this uptake in hazard as a result, not just that increase in asset value and accumulation of wealth.”

He said he has worked with insurers who are able to clean up their books of business, improve on the average vulnerability of the structures that they're insuring on both their residential and commercial side.

Zesty AI uses aerial imagery from companies that collectively cover just short of 100 per cent of the United States.

“We have computer vision scientists on our team who are observing each of these kinds of features. It’s a lot about roofs, a lot about vegetation on the property, the presence of solar panels and swimming pools, anything that they can use to rate and underrate on their side.” He said catastrophe models can be provided with more and highly accurate data.

Mr Wardman’s company, MaxInfo, is a new company developing fresh opportunities for catastrophe model users to “extract Maximum Value” from the cat model data they are already getting. It “equips the re/insurance sector with catastrophe modelling analytics solutions which maximise the information risk modelling can, and should, provide.”

Ms Clark said KCC is involved in second generation models that address the challenges the industry faces with the first generation.

She said secondary perils, such as severe convective storms, winter storms, and wildfires, disruptive model updates and, “major model misses” that included Hurricane Irma and Winter Storm Yuri left low confidence in some of the catastrophe models, even as costs in the modelling industry are rising.

She said KCC hired a whole new “fresh team of PhDs, postdocs, scientists” to develop a new approach, the results for which have been verified by data returned to them by customers.