NAIC to make insurers show AI use in claims and models

• NAIC AI evaluation tool functions nearly finalized
• Regulators scrutinizing both approved and denied claims
• AI model bias also on regulatory radar

U.S. state insurance regulators are advancing an evaluation tool to collect more details of how insurers are using AI, according to insurance compliance experts.

The National Association of Insurance Commissioners (NAIC) AI Systems Evaluation Tool, launched in pilot form in March, is collecting information on what AI systems insurers have in place, how they are being used, how they are governed, how they are monitored to avoid AI drift and how they are performing, according to J.P. Wieske, co-founder and principal of the American InsurTech Council. The council is a group of insurtechs and insurers advancing members' interests concerning regulation.

Wieske, speaking in a recent Insurtech Association webcast, attended NAIC's summer meeting last month. "One of the regulators brought up that they like to also take a look at approved claims, not just denied claims, and how those are looking, because that can certainly impact their solvency," he said. "If you're approving claims that should not be approved, that means that your pricing is going to be off."

With the tool, state regulators are getting a better understanding of how insurers view AI and deploy AI, as well as insurers' relationships with AI service providers, Wieske added. "They're going to be able to see whether or not there's some best practices going on, whether or not the oversight is in fact there, and whether or not there needs to be a little bit more work," he said. The regulators can compare what insurers are doing with AI and whether the results of insurers' use of AI are more or less risky, Wieske said.

State regulators are figuring out what they need to know to determine whether insurers are using AI technology responsibly, according to Scott Harrison, co-founder of the American InsurTech Council. "Regulators want to know that companies know what they're doing," he said. "When it comes to managing their investments, operating a holding company, paying claims, doing their core business processes, regulators need to know for sure that companies are doing things in a compliant way."

Harrison advised startup insurtechs to ask themselves what regulators need to know about their technology. "The regulator has responsibility to understand and ensure that the company is performing competently with respect to this technology," he said.

The state regulators are also looking at whether insurers' AI models are discriminating in any way, according to Harrison. "The concern here is that these models are opaque, and what's behind the model is not necessarily traditional actuarial science," he said. "The regulator might not understand how the company is getting to its answer, and the concern is that either intentionally or unintentionally, the model may be manipulated in such a way to achieve a discriminatory outcome or an unfair outcome, and nobody would know it."

NAIC had planned to pilot its AI Systems Evaluation Tool through September, then update it based on feedback, and present it at its next general meeting in November for adoption.