Main content

The insurance implications of AI are heating up - reviewing the NAIC's Model Bulletin for the use of algorithms and AI

Neil Raden Profile picture for user Neil Raden August 16, 2023
Summary:
The NAIC's highly anticipated Model Bulletin for the use of algorithms, predicative models and AI came out in July, but what are the implications? Though the bulletin is still in draft form, those with a stake in the insurance industry should be prepared.

finger-pointing

In July 2023, the National Association of Insurance Commissioners (NAIC)  released its Model Bulletin, Use of Algorithms, Predictive Models, and Artificial Intelligence Systems by Insurers.

The release of the bulletin is significant, but first, a few facts about the US insurance industry from the Insurance Information Institute:

  • The U.S. insurance industry employed 2.8 million people in 2021, according to the U.S. Department of Labor. Of those, 1.6 million worked for insurance companies, including life and health insurers (911,400 workers), P/C insurers (628,600 workers) and reinsurers (26,900 workers). The remaining 1.2 million people worked for insurance agencies, brokers, and other insurance-related enterprises.
  • U.S. insurance industry net premiums written totaled $1.4 trillion in 2021, with premiums recorded by property/casualty (P/C) insurers accounting for 53 percent, and premiums by life/annuity insurers accounting for 47 percent, according to S&P Global Market Intelligence.
  • P/C insurance consists primarily of auto, homeowners, and commercial insurance. Net premiums written for the sector totaled $715.9 billion in 2021.
  • The life/annuity insurance sector consists of annuities, accident and health, and life insurance. Net premiums written for the sector totaled $635.8 billion in 2021.

Anthony Hayadeb, CEO of Monitor, an AI governance software company, in a recent article, explained nicely why insurance is so central to the economy and our everyday lives and, in particular, how every innovation depends on the insurance industry:

The average person doesn’t realize how consequential and impactful insurance is on our everyday lives. How did that highway get built, that office building get erected, that new medical device get launched, that plane take flight? None of those happen without insurance. From what started as ship merchants pooling dollars to protect each other’s fleet and cargo, insurance is the safety net and enabler of every major project, innovation and industrial revolution. Insurance is what catches us and props us up as individuals during some of our lowest moments—when a family member passes, when a car accident happens, when our house is broken into, when hurricanes devastate communities.

What is the NAIC?

Except for broader US regulations for all industries, such as the Fair Credit Reporting Act, the ACA and others, the insurance industry is largely exempt from federal regulation. Instead, regulation is handled at the state level. All fifty states and six territories have their own insurance commissioners. Some state commissioners are influential and often lead, especially New York.

The NAIC provides some cohesion to the industry, especially the mandatory NAIC Statutory Annual Reports. More broadly, its role is to protect the public interest, promote competitive markets, facilitate the fair and equitable treatment of insurance consumers, and promote the reliability, solvency, and financial solidity of insurance institutions.

Several factors contribute to the restriction of the federal government from directly regulating the insurance industry:

1. Constitutionally Reserved Powers: The U.S. Constitution grants states the authority to regulate commerce within their own borders. Insurance regulation is considered an exercise of this reserved power, and it falls under the state's police powers. The federal government's powers are limited to those specifically enumerated in the Constitution.

2. McCarran-Ferguson Act: In 1945, Congress passed the McCarran-Ferguson Act, which specifically exempts the insurance business from federal antitrust laws and affirms the primary role of the states in regulating insurance. This act essentially allows states to regulate insurance without significant federal interference.

3. Tradition and Historical Practice: Insurance regulation has been traditionally handled by individual states since the 19th century. The states have developed their own regulatory structures, laws, and agencies over time.

4. Diverse Market and Local Variations: The insurance industry varies significantly across different states due to varying local needs and market conditions. State-based regulation allows for more tailored approaches to address these differences.

5. Political Considerations: The insurance industry has historically been a powerful lobbying force at both the state and federal levels. The state-based system allows the industry to exert influence within individual states' regulatory processes.

However, while the federal government is restricted from directly regulating the insurance industry, there are some areas where it does have a role:

1. Federal Agencies: Some federal agencies have oversight roles related to specific aspects of insurance, such as the Federal Insurance Office (FIO) within the U.S. Department of the Treasury, which monitors the insurance industry and provides advice to policymakers.

2. Federal Legislation: Although the regulation of insurance itself is primarily a state matter, federal laws can have an impact on the insurance industry. For example, the Affordable Care Act (ACA) introduced significant reforms in health insurance and the Dodd-Frank Wall Street Reform and Consumer Protection Act brought changes to aspects of the insurance industry, particularly regarding financial stability and systemic risk.

Overall, the state-based regulatory framework has been the dominant approach to insurance regulation in the United States, and any changes to this arrangement would require significant legal and political considerations.

What does the NAIC Bulletin address?

It is just a draft version, subject to comments and revisions, but at first glance, there are some significant points:

First, unlike similar statements from various organizations, governments and otherwise, it clearly expands the scope beyond what is collectively called AI:

Decisions impacting consumers that are made or supported by advanced analytical and computational technologies, including artificial intelligence (AI) systems…specifically with respect to predictive models: the Insurer’s processes and procedures for designing, developing, verifying, deploying, using, and monitoring predictive models, including a description of methods used to detect and address errors or unfair discrimination in the insurance practices resulting from the use of the predictive model.  

The significance of this is including all predictive models, including the inventory of legacy systems. On the one hand, limiting this regulation to just new and current AI technology would overlook a substantial number of systems already in place, but on the other hand, insurers may find the effort in locating and evaluating all of them an impossible burden. I expect a great deal of discussion on this point.

Also, in the above statement is the inclusion of word “consumers.” Unfortunately, this seems to follow the same approach of the established the AI Ethics community and its singular focus on “harms” to people. In fact, the wording limits the scope to “consumers,” even more restrictive than people. Whether referring people or consumers, it seems to overlook the substantial commercial Property and Casualty sector, which is somewhat less-regulated than personal lines, especially in rate filings, but claims processing, solvency, policy forms, standardized reporting and a host of other areas are within the range of NAIC regulation. 

One other item in the bulletin in the draft in the draft is likely to be subject to discussion and revision:

A centralized or federated committee comprised of representatives from all disciplines and units within the Insurer, such as business units, product specialists, actuarial, data science and analytics, compliance, and legal. 

An AI Ethics oversight committee was an early suggestion of commentators on AI Ethics, and was not, in my opinion, informed by understanding how organizations operate in devising digital predictive and proscriptive systems. Such a program would add a level of bureaucracy to the process that would impair the ability of the organization to operate efficiently. Oversight and governance are needed but standing committees have been shown to be ineffective.

My take

Other than some reservations I have in the above, the approach is sound and follows recent efforts by the NIST and FTC to not propose new, sweeping legislation, but rather to provide a framework which I covered here and here. I will follow with subsequent releases. 

Loading
A grey colored placeholder image