Technology snafus don't have to be malicious to cause significant disruption. The CrowdStrike incident in July brought this to light after one faulty software update grounded flights and brought banks and hospital systems offline.

While hacks and scams are among the more eye-catching ways that the increased use of artificial intelligence (AI) can expose businesses, awareness of AI performance failure risks is also beginning to emerge.

“AI performance failure happens when AI-based technology operates correctly but fails to deliver or meet expected performance standards," says Seth Johnson, senior cyber risk and adjusting professional at Charles Taylor. “We're used to putting AI technology into the equation of, 'How can we be compromised from AI-related adversarial attacks?' but, increasingly, we're also including the perspective of, 'What happens when this technology does not achieve its intended purpose, and what are the consequences?'"

One AI performance failure scenario gaining traction today is when “a customer-facing AI program uses software to analyze huge amounts of data to generate written words intended to guide a customer—and while the software and the data may be working together as designed, the output may not be reliable," Johnson says.

The insurance industry has made significant advancements in developing cyber insurance, but “because the primary focus has been addressing risks related to adversarial events, AI performance failures may not fit neatly into many cyber insurance policies," he says. “I've seen many carriers respond differently as they continue to recognize gaps in coverage or areas they may have overlooked."

“Some carriers are finding ways to restrict this coverage if they don't have the appetite for it and others are finding creative ways to enhance it," Johnson continues. “Those who are starting to recognize and embrace coverage related to these types of scenarios are being very specific about the types of AI performance failures they'll cover, often in endorsements."

As far as whether AI performance risk coverage will become more commonly available, “we're still at a crossroad," Johnson says. “Historically, time has aided the industry in making coverage decisions and the industry is still hard at work forecasting the potential scenarios AI performance failures may bring."

“However, AI performance failures will likely become more prevalent with time , and overall, I think we will see the insurance industry increasingly embrace the coverage to support its policyholders," he adds.

As far as what type of clients are most vulnerable to AI performance failure risks, health care or transportation dominate AI risk-related news, but “this is widespread and related incidents occur outside the headlines in all sorts of businesses and industries," Johnson says.

Meanwhile, the insurance industry itself isn't immune to this risk, particularly as AI continues to be adopted by carriers and independent agencies alike. For independent agency owners seeking coverage for their business against performance failures, “it's important to consider cross-coverage implications, which are potentially many, that may respond to the consequences of AI performance failures," Johnson says. “Although there are new coverages that can be very valuable to independent agencies, they are still developing, and we have once again entered a time where a heightened review of existing coverages is necessary."

“Rather than expecting that this type of scenario is covered by existing cyber insurance policies, we may find ourselves relying on coverages like errors & omissions if, for example, AI fails to perform when assisting decision-making or customer support," he adds.

AnneMarie McPherson Spears is IA news editor.