Raising the Risk Level

• Lori Chordas
• October 2017
• 

Inventor and business magnate Elon Musk embraces artificial intelligence. He also fears it.

For years, Musk has been warning of the perils of AI, calling it "the biggest risk we face as a civilization" at a recent National Governors Association meeting.

Insurers, too, see artificial intelligence, which goes by many names such as machine learning, augmented intelligence and others, becoming risky business, but in the form of new liabilities that will impact coverage lines such as auto, product liability, workers' compensation and professional liability.

Insurers themselves don't fear AI. In fact, they're embracing the simulation of human intelligence processes by machines that until recently was merely a figment of the imagination of science fiction writers and the idiosyncrasies of movie characters such as HAL 9000 from 2001: A Space Odyssey and Data from Star Trek.

But AI is far from the stuff of science fiction. AI is here and it's permeating every industry. And that's forcing insurers to think long and hard about how the new norm will change the risk profile.

"We're already seeing that in areas like autonomous vehicles," said J. Stephen Zielezienski, senior vice president and general counsel at the American Insurance Association. "Every state law has its own licensing qualifications for human drivers. But what happens when those drivers are replaced by a computer or machine? That's going to significantly change the nature of auto insurance by fundamentally altering what it means to base insurance premiums on a vehicle operator's driving record."

Changing Drivers

Insurers are starting to devise ways to address the new liabilities that will come from fully autonomous vehicles, along with the transitional risks that will emerge even earlier when both human-driven vehicles and autonomous cars share the roadways.

While frequency and severity of auto claims are expected to decline, costly repairs to autonomous vehicles could come at a hefty price for insurers. The future of the auto insurance market could also be turned on its head if other projections play out. The industry, which currently boasts about $200 billion in premiums, could contract by as much as 60% by 2040 as accident damage payouts and premiums fall, according to consulting firm KPMG.

Also, there are other murky questions still left unanswered about regulation of the technology and who is liable when the technology fails and an accident occurs.

"Insurers are now starting to think about how products will need to change to reflect those exposures, including whether or not personal auto policies now become commercial lines coverages or whether some hybrid product is needed," Zielezienski said. "The answer to some of these questions could change significantly if the traditional automobile sales model in the U.S. shifts away from personal car ownership."

XL Catlin is examining the changing risk profile that comes from driverless vehicles. Last year, the global insurer began partnering with autonomous vehicle software developer Oxbotica to explore the emerging mobile autonomy and technology of driverless vehicles and society's response to their risks.

The U.K. government-funded consortium, known as DRIVEN, hopes to have a small fleet of cars, which will operate with SAE Level 4 autonomy (highly automated self-driving that does not require human driver intervention), travel between London and Oxford autonomously by 2019.

The findings of that project will help insurers better understand how to deal with the risks specific to insurance for autonomous vehicles, said Nancy Bewlay, the casualty global chief underwriting officer at XL Catlin. The findings also will aid Oxbotica in its understanding of what risks are acceptable to underwriters so it can better design technology that can be more readily insured, she said.

In addition to the new auto liabilities, workers' compensation will also feel the effects of AI. A University of Oxford study estimates that 47% of all jobs in the United States could be automated by 2033.

Earlier this year, Japanese insurer Fukoku Mutual Life Insurance announced it was replacing 34 of its employees with IBM's Watson Explorer. The cognitive search and content analysis platform can analyze and interpret data, including unstructured text, images, audio and video, and calculate payouts to policyholders.

The loss of jobs like that will reduce payroll and will decrease workers' comp premiums, said Celent senior analyst Michael Fitzgerald.

There's also some fear that AI could create new risks in the workplace, such as robot-related injuries from programming, installation, maintenance or repair error. Two years ago, a German worker was killed by a robot in a VW auto manufacturing plant.

But there's an upside that comes from automation. Experts say AI allows robots and machines to operate, sometimes for the first time, in dangerous environments, thereby reducing workplace accidents and injuries.

"Also, wearable technology is helping employers gain better insight into when and how employees are working," Bewlay said. "It can help alleviate or lower risk by showing what happened if an employee gets hurt and if the injury occurred while on the job, which has always been an issue in workers' comp," Bewlay said.

AI is also expected to add a new layer to cyberrisk.

Zielezienski said the double-edged sword is that while AI systems can help defend against cyberattacks, they also present new targets for hackers, potentially posing a number of new cybersecurity vulnerabilities for individuals and businesses.

Think about a locksmith who sells video-enabled doorbells that use some type of facial recognition technology, Fitzgerald said.

"If something goes wrong with the installation or someone gets pulled into a legal suit with the manufacturer because an intruder used a 3D facial printout to gain access into a home, that creates new risks," Fitzgerald said. "AI, sensors and technology agents change the risk profile. A doorbell is no longer just a doorbell and a lock is no longer just a lock."

Insurers are starting to evaluate possible liabilities like that and are carefully writing coverages to address those concerns, he said.

"There may be some sort of government role in that, but a more pragmatic way to look at those risks is in the private market, which will evolve as the industry better understands the nature of the relationship between cyber and AI," Fitzgerald said. "That's already happening organically."

Feeling the Heat

AI also opens the door to more complexity in products' liability and potential changes in product liability litigation, said Kevin H. Kelley, CEO of Ironshore Inc.

Advanced manufacturing continues to become more complicated, thereby increasing the likelihood that claims will be filed against not only a product manufacturer but also the companies that enable its technology, he said.

Today, products aren't just made the old-fashioned way, manufacturers are also using robotics or some other means of AI, Kelley said. "That increases the number of potential defendants in litigation, which, historically, has made legal disputes more expensive and protracted. We think that's an area that needs to be addressed."

Changes like that raise a number of questions for insurers. Do current policies cover new risks? Will carriers seek to exclude those new exposures? Are new products needed to address evolving needs and shift risk pools?

Many of those answers are still unclear at the moment, said Will Eustace, executive vice president and head casualty coverage officer at Willis Towers Watson. "But it's something the industry really needs to be thinking about now."

In and Out

While AI is set to create a web of new exposures, it's unclear whether some traditional risks will disappear in its wake.

"I don't know if they'll go away, but workers' comp, for example, might be reduced to the extent manual labor is reduced," Eustace said.

"When automated teller machines came along everyone was worried that tellers' jobs would disappear, but there are actually more tellers today than ever. That's similar to what we'll see with risks caused by AI," he said.

"There is certainly going to be disruption as things change, but hopefully we get to a point where people are performing more value-added, customer-oriented work and AI frees them up from having to do manual tasks," Eustace said.

Potential ramifications of AI and its impact on risk don't stop there.

AI can help carriers model loss scenarios, Eustace said. "Take terrorism, for example. We don't know the frequency and severity of a terrorist attack, but data can tell us things that we as humans can't sift out by simply reading documents.

"When you think about pandemics like Ebola and the 1918 flu, data can now tell us where and when pandemics will hit and their impact," Eustace said. "Is it perfect? No, but it gets more refined as we move along and gives carriers a sense of what could happen and what exposures they have," he said.

"The same goes for cyber," Eustace said. "Data can tell you where a breach is coming from, how large it will be and how often it may hit. Those will be impacts of AI on the industry as well."

Fitzgerald also believes AI will impact the insurance model of avoidance. In commercial lines, for example, AI might reduce frequency of loss because "we'll be able to find previously known patterns that could result in lower frequency, lost work days and accidents. That will definitely affect coverages like workers' comp and business interruption," he said.

AI forces insurers "to reach away from historical processes in order to build a decision-making process that isn't necessarily tested yet," Bewlay said. "It's a new thing for the industry to be able to work in parallel with what we know historically, and move into a process with various data sets and different ways to look at risks."

Insurtech startups have the upper hand in that, she said.

"They're building new ways to analyze insurance risk and they aren't tied to the past," Bewlay said. "They're much more nimble than legacy companies because they're only looking to the future," she said.

Taking Action

The emergence of AI creates a host of potential loss scenarios, each with its own set of unique variables.

Is a cyberattack caused by some form of automation covered in an insured's property policy? Does an equipment manufacturer's product liability coverage insure part of the risk? Is there a business interruption exclusion?

"We can begin to answer those types of questions by observing how exposures are changing and asking clients how AI is impacting, for example, their manufacturing process," Kelley said. "But it's still too early to really know that yet."

Once that begins to show some clarity, "then we need to make sure current contract language is suitable for those exposures and address other changes," he said.

That's where innovation comes into play, XL Catlin's Bewlay said. "We need to look at it not only from an insurance and product perspective but also from an investment standpoint, such as insurtechs focused on AI," she said. XL Catlin's venture capital arm, XL Innovate, has funded a number of industry startups that leverage AI and machine learning.

Education is key, Kelley said. "And when it comes to gearing up for AI, educating clients and staff is now our No. 1 task.

"We want to make sure we know how changes in exposure might affect our own portfolio and ensure that we are not overexposed to what may come down the road," Kelley said.

"A lot of what we do internally is look at claims activity from different sources and make sure we understand how recovery on those claims may occur," Kelley said. "What carriers don't want to do is pick up exposures inadvertently in areas they didn't intend."

The New Normal

This isn't the first time the industry has been promised benefits from AI, Celent's Fitzgerald said.

"Grand pronouncements about what AI can do have come in different cycles over the years, but this time might be different, thanks to the availability of the cloud, mature IT tools and the overabundance of digital data," he said.

"This time around, AI may actually find a way into the core business stream," Fitzgerald said. "If it does, that may impact insurers' risk profiles for both individuals and businesses."

The world has experienced big shifts like that before, Eustace said. The turn of the 19th century brought the economic transformation from agriculture to an industrial revolution and later, there was the move from industrial to technology, he said.

"Now we're seeing a new level of advanced technology," Eustace said. "AI will be disruptive in the short term but it will become very additive to the economy in the long run."

Shifts like that always require some big changes, AIA's Zielezienski said. "The problem is people don't like change.

"The current pace of innovation is challenging our ability to adapt to change and accelerating the need to alter an insurance regulatory system that's been in place since the mid-1940s," he said.

"How do you change functions that have relied on that traditional business model of insurance for decades, and how do you introduce change in a way that won't be disruptive to those who are uncomfortable with it? Regulation has to be seen as a platform for change--a launching point for innovation; otherwise, it will be an obstacle," Zielezienski said.

Right now the biggest hurdle is uncertainty, he said. Uncertainty about what many of the risks are going to be is just part of the new normal in today's AI world.

--------------------

Learn More

Ironshore Insurance Ltd. A.M. Best # 078528

XL Group plc A.M. Best # 052919

For ratings and other financial strength information visit www.ambest.com

By Lori Chordas, senior associate editor, Best's Review: lori.chordas@ambest.com