As artificial intelligence grows in usage, so does the risk, an audience at the Bermuda Captive Conference heard.

An informal poll at the event at the Hamilton Princess & Beach Club found that although the majority of people in the sizeable audience used AI at work and at home, no one had actual AI liability cover.

To add to that, many insurers are now excluding AI from coverage in cyber, commercial general liability, directors and officers and traditional policies.

“There just needs to be an understanding that there is this orphaned AI risk on the parent company balance sheet,” said Troy Dort, director, management consulting at KPMG moderating in a panel called Oh no! Not Another AI Panel!

His solution is to buy AI liability coverage through endorsements from existing carriers or add AI liability into a captive.

“That is as opposed to going out to the standard market and paying the risk load premium to understand the new coverage,” he said.

Panellist Sarah Hopkins, a senior associate at Fenwick who advises companies in the insurtech, fintech, and emerging‑technology sectors, explained that AI is being pulled out of traditional policies because it is a very different risk.

“The amount of risk that it represents is significant to the point that insurers are wanting to pull it out,” she said.

Ms Hopkins said in some cases they are not insuring it at all right now until they can understand it better.

“Many folks, insurers included, don’t understand exactly where the edges are,” she said. “They don’t know how to get their arms around it, and so you see fear seeping into all of this.”

Who needs AI coverage and what that looks like depends on the entity and the purpose, she said.

“When cyber was new on the market most of what we were doing was helping manuscript cyber coverage,” Ms Hopkins said.

Manuscript cyber coverage means creating an insurance policy or endorsement from scratch, rather than using a standard form.

“Now that we are seeing more and more standardisation, we are back at the beginning with AI,” she said. “There is going to be a lot more manuscripting than you see happening on the coverage side.”

Federico Candiolo, consultant senior corporate counsel at ASW Law expects more captives to embrace policies that cater specifically to AI risks.

“Because people are going to be using public tools a lot more and maybe creating different risks for their companies, you are going to see a lot more training,” he said.

He tied that to boards being responsible for the oversight and use of AI, with governance expectations increasing.

Thomas Galbraith, cofounder and CEO of Barkr, expects to see more adoption of both large and small language models in the next year.

“The AI consult models are going to start to do things which fundamentally change the way people operate on their cases,” he said. “We are starting to see this already. There will be models which know you so well, and understand your data so well that they can manage your daily life.”

There have been concerns from some quarters that some functions of the insurance industry could be entirely erased or diminished by AI.

James Wencil, consulting actuary at Pinnacle Actuarial Resources, thought there would always be need for actuaries.

“The technology cannot sign a statement of actuarial opinion,” he said. “It will augment our process and really help us get past the data grind stage and into the decision-making and judgment stage faster.”

However, he said AI has to be used with caution to avoid creating risk or diminishing the value of a partner’s data sets.