How do you insure artificial intelligence?

Picture this. Your client is manufacturing a car that includes software to apply the brakes automatically when approaching slow traffic. No doubt, this software has some built-in artificial intelligence (AI) elements.

When insuring a vehicle manufacturer, it’s not too difficult to spot the potential for critical failure, especially one that could pose risks for consumers. But how does the AI itself come into play? How do insurers pin the risks?

When it comes to insuring the unpredictable, Nick Kidd, director of business insurance at Mitch Insurance, says companies using AI won’t always encounter doomsday scenarios. In fact, he said, insuring AI is more mundane than one might expect.

“I don’t know many tech companies that don’t have AI in them somewhere,” he added.

And many products and services people use every day have AI features built into them.

“Sure, AI has the prospect of leading to some pretty catastrophic losses in a relatively small number of circumstances, but then so do millions of other things,” said Kidd. “[Insurance] has been solving these kinds of very complex risk equations for years.”

Most of the time, AI is encompassed in another form of a client’s coverage and often is not a thing in and of itself to insure.

“It’s very rare someone is just insuring AI,” Kidd noted. “They’re insuring their company and all its exposures, and the reality is AI is usually a component of something bigger.”

That’s because AI is often part of a product or service, or used somewhere in the chain of developing that product or service. For that reason, he said, it can be virtually impossible to just insure the AI component of a product, because it can be difficult to apportion the loss to AI components, or other components, because they work in tandem.

Ruby Rai, cyber practice leader at Marsh Canada, said AI coverage is a technology risk, not just a cyber risk.

“Artificial intelligence is just like any technology,” Rai said. “It’s used differently, and at the end of the day it still might be operated [in collaboration with] humans, so there’s the human error aspect.”

Rai added that AI can become increasingly difficult to monitor as organizations are challenged to anticipate downstream impact. “Organizations need to review the overall impact it has on all aspects of business and to its consumers,” she said.

What can be challenging about insuring AI isn’t the technology itself, but where the technology comes from and who’s involved.

“Is the client buying this artificial intelligence from someone else — so someone else is creating [it] — versus they’re creating it for themselves?” Rai asked. “Most likely, they’re buying it from someone else. So, the organization that is creating that artificial intelligence carries a lot of liability themselves.”

Companies seeking the right coverage need an insurer that understands the maximum exposure artificial intelligence use can create.

“You need to be working with a risk partner that understands that this is not a cookie-cutter [solution],” said Kelly MacDonald, Aon’s regional sales director and senior vice president of commercial risk solutions.

“It becomes a really fulsome collaborative discussion with the client about what risk they want to retain, what risks they could transfer, what their total cost of risk [is],” said Katharine Hall, Aon’s senior vice president and cyber practice leader of commercial risk solutions.