We previously reported on regulators’ increased attention to the use of big data systems, including external consumer data and information sources, algorithms and predictive models. Recent announcements in Colorado, Louisiana and New York addressing the use of not only big data but also more traditional “scoring” models signal a continued focus on this area by state insurance regulators.
The Colorado Division of Insurance has started the stakeholder engagement process to implement Senate Bill 21-169, which was signed into law on July 6, 2021, and directs the Commissioner of Insurance to work with insurers, producers, consumer representatives and other interested parties before adopting regulations requiring insurers to demonstrate that their use of big data does not unfairly discriminate against a protected consumer class. This process will initially focus on life insurance underwriting practices, with the first meeting scheduled for February 7, 2023, and later expand to auto, health, marketing and claims practices. All interested stakeholders are invited to participate in the upcoming session and can register here.
On January 25, 2023, the Louisiana Department of Insurance issued Bulletin 2023-01 to “[a]ll authorized insurers and surplus lines insurers” regarding the use of crime statistics in underwriting and rating property insurance policies in St. Francisville and West Feliciana Parish. Certain insurers have attributed property insurance declinations in these areas to “negative crime scores provided by third-party underwriting sources” that reflect Federal Bureau of Investigation (FBI) crime statistics on a county/parish basis. St. Francisville is a small community with a population of less than 2,000, and West Feliciana Parish’s population of 15,000 includes the 5,000 inmates who reside in the Louisiana State Penitentiary (also known as Angola). Because crimes within Angola are included in the FBI crime database as having occurred in St. Francisville/West Feliciana Parish, third-party data sources calculating crime scores using the FBI crime database report an “F” rating, which adversely distorts the crime scores for these areas. Excluding inmate-on-inmate crimes within Angola, St. Francisville/West Feliciana Parish have very low crime rates. The Bulletin “urges and requests” insurers to “properly evaluate and underwrite properties located in St. Francisville and West Feliciana Parish considering the information provided above” and “abide by the legislative intent of La. R.S. 22:41 that policyholders shall have the right to be treated fairly.”
Following its 2019 publication of Insurance Circular Letter No. 1 addressing big data in underwriting life insurance, the New York Department of Financial Services (NYDFS) wrote to various industry stakeholders last week about the next phase of its efforts to prevent unlawful discrimination from the use of big data and other modeling approaches. NYDFS has engaged consultancy Fairplay-Sustain Solutions LLC to help it “better understand [its] continued role in the adoption of artificial intelligence and machine learning (AI/ML) in the insurance sector, including the potential benefits and harms that can arise from the use of AI/ML-derived underwriting models.” Fairplay-Sustain will conduct a series of stakeholder interviews and conversations with life insurers, law firms, other states/jurisdictions, consumer advocates and data providers. NYDFS expressed its hope that stakeholders will share candid feedback and insights. NYDFS’s outreach states that Fairplay-Sustain is under a non-disclosure agreement with NYDFS and that anything shared, “including confidential information and/or future plans, will not be shared, disclosed, or used in any context outside of this engagement by any party.”
Engagement with regulators in Colorado and New York may become important precedent in the industry’s ongoing innovation efforts as regulators continue to find their footing with respect to big data, AI/ML, algorithms and other information sources. The Louisiana Bulletin illustrates a big data system or predictive model, and its resulting report or score (i.e., the quality of output) is only as reliable (and lawful) as the quality of the input. Further, systems or models applying hyper-localization or other more individualized risk factors could benefit protected classes and otherwise help prevent unfair discrimination in underwriting and rating by identifying outliers in a dataset. These themes should be prominent in the upcoming regulator-industry dialogues in Colorado and New York and in the broader discussion going forward.
Fairplay-Sustain will interview members of the McDermott Insurance team in late February as part of NYDFS’s study. Please reach out to one of the authors of this article or your regular McDermott lawyer if you would like them to convey any insights, comments, questions or other information.