Locke Lord QuickStudy: Colorado Exposes Draft Life Insurance ‎AI Data Testing Regulations for Unfair ‎Discrimination

On September 27, 2023, the Colorado Division of Insurance (the “Division”) exposed a new proposed regulation entitled “Concerning Quantitative Testing of External Consumer Data and Information Sources, Algorithms, and Predictive Models Used for Life Insurance Underwriting for Unfairly Discriminatory Outcomes” (“Testing Regulation”). The Testing Regulation is being developed in connection with Colorado SB 21-169, enacted in June 2021. On September 21, 2023, the Division previously adopted its first regulation implementing Colorado SB 21-169, which addressed requirements for Governance and Risk Management Frameworks. The newest proposed regulation addresses the quantitative testing requirements for life insurers that use ECDIS to ensure that their use is not unfairly discriminatory based upon race and ethnicity.

The proposed Testing Regulation applies only to life insurers selling life products. The quantitative testing is intended to determine if underwriting decisions are unfairly discriminatory based on the race or ethnicity of the applicants utilizing “BIFSG and the insureds’ or proposed insureds’ name and geolocation information.”  The proposed Testing Regulation defines “BIFSG” to mean “the statistical methodology developed by the RAND corporation for estimating race and ethnicity.”

Initially, the test should include application data for the period ending December 31, 2023, and thereafter through December 31^st of the preceding year. The proposed Testing Regulation requires three different types of testing:

Application Approval Decision Testing

Using the BIFSG estimated race and ethnicity of proposed insureds and the following methodology, insurers shall calculate whether Hispanic, Black, and API proposed insureds are disapproved at a statistically significant different rate relative to White applicants for whom the insurer, or a third party acting on behalf of the insurer, used ECDIS, or an algorithm or predictive model that used ECDIS, in the underwriting decision-making process.

* * *

Determine if there is a statistically significant difference in approval rates for each BIFSG estimated race or ethnicity variable as indicated by a p-value of less than .05.

a. If there is not a statistically significant difference in approval rates, no further testing is required.

b. If there is a statistically significant difference in approval rates, the insurer shall determine whether the difference in approval rates is five (5) percentage points or greater as indicated by the marginal effects value of each BIFSG estimated race or ethnicity variable.

i. If the difference in approval rates is less than five (5) percentage points, no further testing is required.

ii. If the difference in approval rates is five (5) percentage points or greater, further testing is required . . .

Testing Regulation, § 6.

Premium Rate Testing

Using the insureds’ BIFSG estimated race and ethnicity, insurers shall determine if there is a statistically significant difference in the premium rate per $1,000 of face amount for policies issued to Hispanic, Black, and API insureds relative to White insureds for whom the insurer, or a third party acting on behalf of the insurer, used ECDIS, or an algorithm or predictive model that used ECDIS, in the underwriting decision-making process.

* * *

Determine if there is a statistically significant difference in the premium rate per $1,000 of face amount for each BIFSG estimated race or ethnicity variable as indicated by a p-value of less than .05.

a. If there is not a statistically significant difference in premium rate per $1,000 of face amount, no further testing is required.

b. If there is a statistically significant difference in premium rate per $1,000 of face amount, determine whether the premium rate per $1,000 of face amount is at least 5% more than the average premium rate per $1,000 for all policies.

i. If the difference in premium rate per $1,000 of face amount is less than 5%, no further testing is required.

ii. If the difference in premium rate per $1,000 of face amount is 5% or greater, further testing is required . . .

Testing Regulation, § 7.

Variable Testing

Variable Testing is required to identify the specific variable(s) contributing to the difference in first two types of testing where the testing results in a difference of 5 percentage points or greater.  See Testing Regulation, § 8.

Insurers must provide a summary report of the testing results as described in the proposed regulation on April 1, 2024, and annually thereafter. There is an enforcement mechanism for noncompliance in accordance with existing remedies. See Testing Regulation, § 9.

Written informal comments on the proposed regulation must be submitted by October 26, 2023 to DORA_INS_RulesandRecords@state.co.us. A hearing has been scheduled for October 19, 2023 at 12 pm central.