How insurers are grappling with silent AI

Brokers and policyholders are pushing to obtain cover for risks related to artificial intelligence (AI) in products lines such as cyber, directors’ and officers’ (D&O) liability and professional indemnity, the Insurance Day podcast has heard.

Jonathan Hopkins, an associate at law firm DAC Beachcroft, said brokers are increasingly asking for specific policy language affirming AI-related coverages rather than leaving those risks silent.

But while some insurers have been adding affirmative cover to their wordings, others have been seeking to apply broad exclusions, Hopkins told the latest edition of the Insurance Day podcast.

“Policyholders are looking for specific language that affirms AI-related risks, rather than leaving it silent,” Hopkins said. “That’s the starting point where questions are arising. But there are insurers out there trying to innovate and affirm it as a benefit to policyholders.”

The issue of silent AI – unintended or non-affirmative exposure to AI-related losses – immediately draws comparisons to the issue of silent cyber that plagued the early years of the cyber market. Cyber under­writers first looked to exclude unwanted cyber risk before building exposure back in through affirmative cover, but AI cover appears to be developing differently.

While regulator pressure and market initiatives including from Lloyd’s pushed insurers to exclude silent cyber, these push factors have been largely absent in AI risk, Hopkins said.

“Unlike cyber, which introduces a completely new type of exposure, AI often amplifies traditional risks. That means it would be difficult, in my view, for an insurer to apply broad exclusions for AI risks across all of their policies,” he said.

This does not mean insurers have not tried. Since the middle of this year, at least three insurers have applied “very broad” AI exclusions across multiple lines including D&O, errors and omissions and fiduciary liability insurance products.

But insurers are also increasingly building affirmative cover into their existing wordings, Hopkins said, pointing out at least one London market insurer has updated its technology professional indemnity wording to specifically include AI risks. On top of this, several standalone AI policies have been launched, mostly providing cover for liabilities arising from performance failures and regulatory exposure, but also for product recalls and crisis management.

“Over the past 12 months or so I’ve seen a number of endorsements and have worked with a number of insurers in the market seeking to affirm AI risk on renewal, not just in cyber but also in the D&O and tech PI wording space [which has] seen the majority of this activity,” Hopkins said.