Few things thrill baseball fans more than opening day. They come with a mix of excitement and trepidation. For many of them, it’s the first day they get to see their favorite team showcase its talents, and their first glimpse at how well their favorite team will measure up to — in some cases lofty — expectations. Attendees and observers alike of the NAIC Spring National Meeting undoubtedly harbored a similar mix of excitement and trepidation when they gathered to hear the Innovation, Cybersecurity, and Technology (H) Committee and the Big Data and Artificial Intelligence Working Group report on current initiatives concerning the development of regulatory frameworks for the oversight of the use of “big data,” artificial intelligence (AI), and machine learning (ML) technologies by insurance companies.

Innovation, Cybersecurity, and Technology (H) Committee

The committee led off with an update on the proposed draft model bulletin for regulatory guidance on big data and AI/ML. Commissioner Kathleen Birrane stepped up to the plate to report that the committee is developing a principles-based draft model bulletin that articulates a high-level view of AI/ML. Specifically, Birrane noted there was strong agreement among committee members to: (1) place the responsibilities for vendors on licensees rather than on the vendors themselves and (2) address validation testing. The model bulletin is expected to contain four sections: introduction, definitions, regulatory expectations, and regulatory oversight/examination. The draft model bulletin is expected this summer.

Next to the plate was Commissioner Michael Conway, who called on pinch hitter Jason Lapham, the big data and AI policy director for the Colorado Division of Insurance, to discuss Colorado’s proposed algorithm and predictive model governance regulation. The division is reviewing comments to the draft regulation, which was exposed for public comment on February 1, 2023. The division plans to work through the comments and make adjustments as deemed appropriate, and will expose a revised draft regulation for further discussion.

Consumer representative Birny Birnbaum attempted a “brushback” of the draft regulation by expressing concern that the numerous reports required by the draft regulation would exceed the division’s ability to review them. But, Conway brushed off that concern by stating that the division has options for expanding its roster.

At this point, Commissioner Amy Beard also entered the game to propose the use of independent data sets to test for bias in models. Beard noted that, while still in the early stages, there is a desire for community engagement on this approach.

Big Data and Artificial Intelligence Working Group

Commissioner Kevin Gaffney led off for the working group, reporting on the status of the life insurance AI/ML survey, a draft of which appears on the NAIC website. The survey includes various questions on how, and in what capacity, insurers are using AI/ML, as well as frequently asked questions and definitions to assist companies in their responses. The survey is expected to be finalized by the end of the month. On May 1, formal call letters will be distributed to 192 life insurers in 14 states. Insurers are expected to respond by May 31. The survey will be issued to: (1) insurers with more than $250 million in premiums on all individual policies in 2021; (2) term writers that have issued policies on more than 10,000 lives; or (3) a specifically selected insurtech company.

Hitting cleanup, Commissioner Doug Ommen reported on comments received on the framework and draft questions for regulators on data and models used by insurance companies. Of the concerns raised in the comments to the draft questions, the following were notable:

• The scope of the questions and the need for a limited, principles-based approach to encourage innovation without overregulation, including redundant regulations.
• The compliance costs for smaller companies and vendors.
• The impact of making insurers responsible for third-party vendors.
• The need to safeguard vendors’ proprietary information.
• Methods for testing data and models.

In light of the comments received, the working group is expected to produce a revised draft of questions for regulators by the end of May.

So will all of these initiatives play up to their expectations? Only time will tell, but for fans of these initiatives, the season is off to an exciting start.