- AI-related claims mostly handled through existing types of insurance
- Some insurers exclude coverage of risks caused by AI
- Affirmative coverage in demand
Silent AI — meaning AI embedded into existing products, services and processes — creates hidden exposures for insurers that may be completely unaware of the risks, according to insurance technology, legal and underwriting experts.
The insurance industry is figuring out whether to address those risks in existing coverage or exclude them, the experts said.
For now, claims for AI-related losses will be made against existing coverages, even as AI redefines risks, said Hayley Budd, head of innovation at Atrium, a U.K.-based underwriting firm, who spoke in a recent
"Whether that is booking a table for dinner or a holiday or authorizing a payment, purchasing goods, making employment decisions, or even getting to a point of handling claims and paying out claims without a human being involved, the liability is switching from an output to what the AI actually does," Budd said.
Exposures to silent AI elements can depend on who is insured, according to Budd.
"A base model provider is going to have a very different exposure to an e-commerce company or platform just deploying off-the-shelf chatbots that they bought from a vendor," she said. "Even if the technology is the same underneath, the liability profile is going to change depending on what you're doing with the AI. AI liability isn't like one definition of one risk. It's a collection of liabilities that arise from different points, different uses, different people."
The response to silent AI risks does not necessarily need to be a change to "affirmative" coverage that explicitly names AI risks in insurance policy language, according to James Cooper, partner at Norton Rose Fulbright LLP.
"The liability is the same as it always has been," he said. "Is it negligence? Whose fault is it? Can we pass liability on to somebody else?"
On the contrary, commercial policies for financial institutions and directors and officers (D&O) coverage are starting to exclude coverage of AI risks in their language, according to Cooper. But excluding or including AI in insurance coverage creates an issue of identifying whether AI is the cause or "bad actor" behind a loss, he added.
"When machines can do things without a human's interaction, how will first-party policies be able to respond to that?" Cooper said.
While AI risks are hard to define, there is still demand for affirmative coverage, according to Nick Gibbs, class leader at Apollo, a Skyward Group insurance platform operated through Lloyd's of London.
"It remains in the background, against a background environment of people cautiously monitoring it," Gibbs said. "There needs to be more activity in the market for there to be greater appetite around it."
Some Lloyd's syndicate insurers, such as Miller and Testudo, offer dedicated AI liability products, and U.S. MGAs are offering similar products, Budd pointed out. These are limited in scope, she said.
"Where coverage exists, it's often relatively modest limits with hard aggregate caps on it that aren't necessarily going to match the client's exposure," she said. "For some, those limits today may be sufficient for how they are using AI. As it becomes more embedded in critical business processes or through deployment of large-scale agent systems, the market is going to need significantly more capacity than what is currently available, both from a primary insurance and a reinsurance standpoint as well."
