Knowledge Base Regulation Tracker Standards & Frameworks OSS Directory AI Incident Database Research Paper Library Case Studies Society Articles News Jobs About Charter Sign inBecome a Member
Home/Regulation Tracker/Colorado SB 21-169 — unfair discrimination in insurance, and Regulation 10-1-1

Colorado SB 21-169 — unfair discrimination in insurance, and Regulation 10-1-1

The first US law requiring insurers to test AI and external data for unfair discrimination, implemented through Division of Insurance regulations that mandate governance frameworks and quantitative bias testing for life insurers.

coloradoinsurancealgorithmic discrimination

Overview of Colorado SB 21-169 and Its Regulatory Framework

Colorado Senate Bill 21-169 (SB 21-169) establishes legal protections against unfair discrimination in insurance practices, specifically targeting the use of external consumer data, algorithms, and predictive models. Enacted on July 6, 2021, this legislation marks the first of its kind in the United States to explicitly require insurers to assess and mitigate discriminatory impacts in AI-driven underwriting and risk assessment processes. The law applies to all insurance practices that rely on external data sources, with a focus on prohibited discriminatory factors such as race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression. The implementation of SB 21-169 is governed by regulations developed by the Colorado Division of Insurance, which outline compliance requirements for insurers.

The legislation was signed into law by Governor Jared Polis on July 6, 2021, with provisions mandating that regulations adopted by the commissioner of insurance would not take effect before January 1, 2023.

The law’s primary objective is to prevent insurers from using external data sources, algorithms, or predictive models that disproportionately disadvantage individuals based on protected characteristics. It achieves this by requiring insurers to demonstrate compliance through rigorous testing and governance frameworks. The bill explicitly prohibits discrimination in insurance practices, whether arising from human decision-making or automated systems. This distinction underscores the growing recognition of algorithmic bias as a critical regulatory concern in the insurance industry.

Key Provisions and Compliance Requirements

SB 21-169 imposes several mandatory requirements on insurers to ensure fairness in the use of external data sources. These include the obligation to disclose information about the data sources used in developing algorithms and predictive models for insurance underwriting and risk assessment.

Central to the law’s compliance framework is the mandate for insurers to establish and maintain a risk management framework. This framework must be designed to identify, to the extent practicable, whether the use of external data sources results in unfair discrimination against individuals based on protected characteristics.

The regulations adopted by the commissioner must also include provisions for remedying unfairly discriminatory impacts in external data sources. Insurers are granted a reasonable period to address identified issues, though the specific duration is determined by the commissioner’s rules. To streamline compliance, the law permits insurers to use external data sources that have already been assessed by the Division of Insurance and found to be non-discriminatory. The commissioner retains the authority to examine and investigate an insurer’s use of external data sources at any time, ensuring ongoing compliance with the law’s provisions.

Scope, Exemptions, and Regulatory Oversight

While SB 21-169 applies broadly to insurance practices involving external data sources, certain exemptions are specified. Title insurance, bonds executed by qualified surety companies, and insurers of exempt commercial policyholders are not subject to the law’s requirements. However, the law’s focus on protecting consumers from discriminatory practices remains a cornerstone of its design, even within the exempted categories.

The Division of Insurance is tasked with overseeing the implementation of SB 21-169, including the adoption of rules that operationalize the law’s requirements. The commissioner is required to conduct a stakeholder process to develop these rules, incorporating input from insurers, consumer advocates, and other relevant parties. The final rules must address specific types of insurance and practices, outlining the methods insurers can use to demonstrate compliance with the law’s anti-discrimination provisions. This collaborative approach ensures that the regulations are practical, feasible, and aligned with industry standards while safeguarding consumer interests.

To enhance transparency and accountability, the Division of Insurance must include specific information in its annual SMART Government Act report to legislative committees. This report must detail any rules adopted under SB 21-169, changes in insurance rates attributable to the law’s prohibitions, and a summary of the stakeholder process. By making this information publicly available, the law promotes greater scrutiny of its implementation and allows stakeholders to assess its impact on the insurance market.

The confidentiality of proprietary information submitted by insurers is protected under SB 21-169. Documents, materials, and other information obtained by the Division of Insurance in the course of compliance reviews are recognized as containing trade secrets. However, the commissioner may use this information in regulatory or legal actions and may disclose it in aggregated or de-identified formats. This balance between confidentiality and transparency ensures that insurers can share sensitive data without fear of competitive disadvantage, while regulators retain the tools necessary to enforce the law effectively.

Required actions

  • Insurers may not use external consumer data, algorithms or predictive models that unfairly discriminate on protected grounds - Life insurers must adopt a governance and risk-management framework for such models and file annual reports and quantitative bias tests with the Division

Sources

Contribute

Spot an error, or know a record we should add? Members can propose additions and submit corrections.