From the use of neural networks that identify fraudulent claims, to the adoption of chatbots that help respond to policyholder questions, artificial intelligence has the potential to transform the insurance industry by automating and streamlining a wide range of insurance processes. AI, however, is not without regulatory risk, and the National Association of Insurance Commissioners (NAIC) recently formed the Artificial Intelligence (EX) Working Group to examine the best means to regulate this rapidly developing technology.
AI often relies on machine learning to learn and improve on tasks that the AI has not specifically been programmed to do. This raises a number of regulatory issues as regulators are concerned that the AI software represents a black box in which they have no insight into how the AI arrives at its decisions. For instance, insurtechs often need prior regulatory approval to use their data models to underwrite insurance. It is more difficult for regulators to scrutinize AI-based underwriting models where the AI decisions can change over time after the model has been approved by the regulators. Without proper regulatory scrutiny the AI could engage in unfair discrimination against protected classes or fail to protect a customer’s data privacy. Consumers may also have no means to dispute a decision, such as the rejection of a claim, made by the AI since they lack the necessary information to do so.
To help address these concerns, the AI working group published a draft Principles on Artificial Intelligence at the NAIC 2019 Fall National Meeting this past December. Principles include transparency, accountability and security. Under these principles, insurance companies are responsible for understanding how their AI system uses its data to arrive at a decision and ensuring that the AI system is in compliance with applicable law. The principles are designed to represent a high-level overview that the working group will use as a basis for developing future AI-specific regulation. The working group is currently soliciting feedback on its draft principles and hopes to present a revised draft for the 2020 Spring National Meeting in March.