Insurers grapple with AI ethics and regulation
With U.S. and Europe both weighing in on insurers’ use of AI, insurers and insurtechs are considering the ethics of AI applications in their business.
In the U.S., NAIC, the association of state regulators, adopted a model for AI use in December 2023, with 17 states now using that model and four more using their own guidelines. In July, the European Parliament passed its AI Act, which sets a legal framework for the use of AI.
Jennifer Kyung, vice president, property and casualty underwriting at USAA.
“AI can be a great — it has been a great help, and will be a great help going forward, to help speed up and gain more insights for the underwriter, as opposed to a black box that’s not explainable sitting off to the side,” said Jennifer Kyung, vice president, property and casualty underwriting at USAA, a panelist on a webcast hosted by the Insurance Information Institute. “It’s really, really important for us to make sure that what we’re looking at is explainable and traceable.”
USAA has instituted model risk management. MMG Insurance, founded in 1897, serves about 200,000 personal and commercial lines policyholders and is working on more use cases for AI technology, including data management, according to Matthew McHatten, president and CEO, who also spoke on the webcast.
Matthew McHatten, president and CEO, MMG Insurance
“The ability to gather data that’s currently not useful or unstructured will become a factor in price and process,” he said. “As we think about this and the reputational risk that’s associated with it, how do we guide expectations? We talk to consumers and regulators, and it’s important to reverse engineer and be proactive and transparent about it. We have to acknowledge how difficult this is going to be to regulate.”
McHatten points to three areas where AI has potential for insurers – reducing friction in transactions, improving customer experience and improving on data and processes.
In insurance and underwriting transactions, medical and legal documents, which are read-only, require a lot of handling time and back-and-forth with policyholders. Customers will hold insurers to the same standard they see from other industries, McHatten said.
“It’s a better opportunity with having the ability for AI to sit beside them, to have more customer knowledge at point of transaction, which leads to a personal approach and positive outcomes,” he said. “There’s significant potential there for consistency, accuracy and better customer experience.”
Moreover, faster response time and proactive attention to customer needs are areas of great promise for AI, according to McHatten. Lastly, harnessing data keeps insurance products “competitive and affordable,” he said, along with reducing risk. “All the information that resides in a claim file could make us better and more effective in how we look at risk. That’s just one example of uncorrelated behavioral data that’s sitting there that we’ll likely be able to harness to provide more competitive products in the future.”
Iris Devriese, underwriter and AI liability lead, Munich Re.
Munich Re, for its part, has both a quantitative and qualitative approach to AI, according to Iris Devriese, underwriter and AI liability lead at the reinsurer. Munich Re uses a model and use case-specific approach, she said. Quantitatively, the firm considers how often AI makes mistakes and what happens if it does. Qualitatively, performance metrics and monitoring, along with bias and fairness audits can help management and produce robust governance, according to Devriese.
“We have a very holistic and in-depth approach to each and every model that goes a little bit further than the standards that are in the market,” she said. Munich Re itself offers an insurance-backed performance guarantee for AI vendors.
Mike Fitzgerald, an advisory industry consultant for insurance at SAS.
NAIC’s AI oversight model includes provisions to control for discriminatory bias, observes Mike Fitzgerald, an advisory industry consultant for insurance at SAS, a data analytics services provider. Generally, he added, regulators will monitor insurers’ building and use of AI systems, as well as their use of third-party AI systems. “The bottom line is this, it doesn’t matter where you write insurance on the earth, this AI regulation is going to touch you somehow,” he said.
