The meteoric rise of artificial intelligence across numerous industries has led insurers, brokers, lawyers and others to pause and consider what new risks and exposures the developing uses of the technology may create. 

While generative AI and its use in “deepfakes” are grabbing attention, commercial insurance claims for losses related to the emerging technology have yet to reach the critical mass necessary to spur insurers to adjust policy language or issue widespread exclusions. 

Change has nonetheless begun, as governments move to develop parameters for the new technology (see related story), and at least one company has introduced an affirmative AI coverage endorsement.

“It will take time for the market to mature on these points for the exposures to be identified,” said Julian Miller, London-based partner at DAC Beachcroft LLP, who has worked on policy wordings for insurers. So far, he has been asked only once to add an AI exclusion to a policy wording.

“From a technological perspective, I’m seeing new nuances to existing categories of risks. I’m not seeing any completely new categories of insurable risks that have surfaced because of generative AI as a technology, at least not so far,” said Jaymin Kim, Toronto-based senior vice president, cyber risk practice, for Marsh LLC. 

Ms. Kim leads emerging technologies work within the global cyber practice at Marsh in a role “to assess whether there’s any net new categories and insurable risks that are surfacing with emerging technologies … not limited to artificial intelligence,” she said.

Over the past 15 months or so, during which high-profile AI technologies have been introduced, “companies across virtually every industry have been reaching out to talk about AI,” the vast majority of which has to do with generative AI, computational systems that run on deep learning techniques intended to create original content, she said.

AI has become a topic of discussion among brokers and policyholders. 

“We make sure it is front and center in all of our conversations, particularly if there is a current cyber renewal,” said Nadia Hoyte, New York-based national cyber practice leader for USI Insurance Services LLC.

Exposures

Organizations are evaluating how AI may affect coverages or trigger claims.

Bob Wice, West Hartford, Connecticut-based head of underwriting management, cyber and tech, at Beazley PLC, said his group writes standalone cyber insurance policies that cover privacy, liability, breach response costs, business interruption and security events, as well as technology errors and omissions policies and media liability policies. 

“That mix … needs to be discussed in the context of what additional exposure should we be thinking about affirmatively covering. Should we be thinking about whether there’s additional exposure that we’re facing that we hadn’t faced before, and how artificial intelligence and generative AI really plays into all of that?” he said. 

“From a coverage point of view, at this point, when you think of generative AI, it’s a tool, not a new form of entity or existence,” said Elisabeth Case, Chicago-based global product manager, cyber, for Liberty Mutual Insurance Co. 

AI-related claims may fall under property/casualty or specialty lines, sources said.

“The types of claims that are likely to come out of the use of AI tools are claims that are covered under existing policies,” said Marshall Gilinsky, a shareholder in Boston and New York for policyholder law firm Anderson Kill P.C., who practices in the firm’s insurance recovery and commercial litigation departments.

“It goes back to the exposures and the risks. They’re already there; it’s just a matter of whether the use of AI is going to make them more prevalent and more efficient, and more numerous,” said insurer attorney Meghan Dalton, a Chicago-based partner at Clyde & Co.

“It is important to note that AI may not introduce new risks, but it changes the frequency and severity of risks that are already in place in ways that we don’t yet understand,” said Matt Harrison, London-based executive director, casualty, for Gallagher Re, the reinsurance unit of Arthur J. Gallagher & Co.

For example, a doctor using AI may be better at diagnosing 95% of diseases, but due to potential biases in training data may chronically under-diagnose or not recognize 5% of diseases. “Misdiagnosis was always a risk. AI just changed it,” Mr. Harrison said.

Michelle Fesi, technical director of special lines at Schaumburg, Illinois-based Zurich North America, said AI training bias can lead to potential exposures.

“Bias is a concern. Language models trained on biased datasets can perpetuate prejudices related to gender, race, religion and other social factors. This can have significant impacts on decision-making processes and customer experiences, potentially leading to legal or reputational consequences,” she said.

Seeking clarity

The industry appears to have started efforts to add contract clarity, according to sources.

“Policies do need to clearly articulate whether this type of risk is included or not,” said Michelle Chia, New York-based chief underwriting officer, cyber, Americas, for Axa XL, a unit of Axa SA. Axa XL is reviewing its insurance policy language with an eye toward AI, she said.

One early mover is cyber managing general agent Coalition Inc., which in March introduced an artificial intelligence affirmative endorsement to clarify what is covered by its U.S. surplus lines and Canadian cyber insurance policies.

The endorsement has seen substantial uptake, said Tiago Henriques, Zurich-based vice president of research for Coalition, with one agency that places coverage with the MGA requesting it be added to all its Coalition policyholders’ policies. 

Coalition introduced the endorsement because it saw a rising influence of AI in some of its claims reviews of cyber incidents, Mr. Henriques said. “We were starting to see those phishing emails increase in quality, and we were starting to see even in some situations deepfakes being used as well,” he said.

The endorsement was initially drafted in mid-2022 and Coalition then consulted with insurers, cybersecurity providers and others to hone the language, he said.

As the commercial insurance sector moves to accommodate artificial intelligence, the new technology appears to be growing in acceptance and penetration among the public.

A recent study by data analytics company Insurity LLC showed that 35% of consumers favor the use of AI in fraud detection in the commercial property/casualty insurance sector, with 32% in favor of AI use to support delivering personalized products and promotions, and 24% in favor of AI in customer service.

The report also showed some reluctance, however, as 50% of consumers said they were against the use of AI in claims management, and 45% opposed AI’s use in underwriting policies.