Written and maintained by CASRAI Editorial Board
Published
On this page
Colorado Senate Bill 21-169 (SB 21-169), enacted on July 6, 2021 and codified at Colorado Revised Statutes (C.R.S.) § 10-3-1104.9, is a state insurance statute that prohibits insurers from using External Consumer Data and Information Sources (ECDIS), as well as algorithms and predictive models that use ECDIS, in any insurance practice that unfairly discriminates against consumers based on race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression. Unlike general consumer AI statutes such as the Colorado AI Act (SB 24-205), SB 21-169 specifically directs the Colorado Commissioner of Insurance to promulgate binding, line-of-business rules—beginning with 3 CCR 702-10 (Regulation 10-1-1)—requiring insurers to establish board-level AI governance frameworks, maintain comprehensive ECDIS and model inventories, and execute statistical bias testing across underwriting, pricing, tiering, and claims administration.
Ask CASRAI · free to tryQuestion about this guide? Ask CASRAI answers with linked sources. 2 free questions a day, no account.
What Is Colorado SB 21-169? Statutory Scope and Key Definitions
Codified within Part 11 of Article 3 of Title 10 of the Colorado Revised Statutes (Unfair Competition and Deceptive Practices), C.R.S. § 10-3-1104.9 addresses the actuarial shift from traditional, directly disclosed underwriting variables toward third-party behavioral, lifestyle, and digital-footprint data feeds. The statute establishes four foundational definitions that govern AI in insurance underwriting and risk scoring:
- External Consumer Data and Information Source (ECDIS) — C.R.S. § 10-3-1104.9(8)(b): Any data or information source used by an insurer to supplement or replace traditional underwriting factors or to establish lifestyle indicators that are used in insurance practices. The statute explicitly lists credit scores, social media habits, locations, purchasing habits, home ownership, educational attainment, occupation, licensures, civil judgments, and court records, while authorizing the Commissioner to expand the definition by rule as new data modalities (such as consumer wearable telemetry, prescription history aggregators, and facial-analytics estimators) emerge.
- Algorithm — C.R.S. § 10-3-1104.9(8)(a): A computational or machine-learning process that informs human decision-making in insurance practices.
- Predictive Model — C.R.S. § 10-3-1104.9(8)(d): A process of using mathematical and computational methods that examine current and historical data sets for underlying patterns and calculate the probability of an outcome.
- Unfairly Discriminate / Unfair Discrimination — C.R.S. § 10-3-1104.9(8)(e): Includes the use of one or more ECDIS, as well as algorithms or predictive models using ECDIS, that have a correlation to race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression, and that result in a disproportionately negative outcome for such protected groups, which negative outcome exceeds the reasonable correlation to the underlying insurance risk that the insurer is able to demonstrate.
Traditional Underwriting Data vs. ECDIS Under C.R.S. § 10-3-1104.9
A recurring compliance question for life, disability, and health insurers is where traditional actuarial records end and ECDIS begins. Under C.R.S. § 10-3-1104.9(5), traditional medical records, motor vehicle reports, or application disclosures provided directly by the applicant are not automatically classified as ECDIS when evaluated in standard actuarial tables; however, when third-party data brokers aggregate, score, or feed consumer attributes into an automated predictive model, the model falls under CDOI’s ECDIS oversight.
| Statutory / Regulatory Construct | Citation | Core Requirement or Scope | Exemptions & Statutory Boundaries |
|---|---|---|---|
| Prohibition on Unfair Discrimination | C.R.S. § 10-3-1104.9(1)(a)–(b) | Insurers may not use any ECDIS, algorithm, or predictive model using ECDIS in a manner that unfairly discriminates based on the 9 enumerated protected classes. | Does not prohibit actuarial differentiation when the insurer demonstrates reasonable correlation to underlying risk and absence of disproportionate protected-class penalty. |
| Stakeholder Rulemaking Process | C.R.S. § 10-3-1104.9(2)–(3) | Commissioner must adopt rules by insurance line (life, auto, health, homeowner) only after convening public stakeholder sessions with insurers, actuaries, and consumer advocates. | Rules are promulgated line-by-line rather than through a single blanket mandate across all insurance products at once. |
| Insurer Compliance Submissions | C.R.S. § 10-3-1104.9(3)(b) | Requires insurers to provide a description of the ECDIS used, how the data is used, and the manner in which the insurer tests and remedies unfair discrimination. | Proprietary algorithmic source code and trade secrets submitted to CDOI are confidential and exempt from the Colorado Open Records Act (CORA) under § 10-3-1104.9(4). |
| Title Insurance & Bonds Exclusion | C.R.S. § 10-3-1104.9(7) | Statute applies to life, health, property, and casualty lines regulated under Title 10. | Explicitly excludes title insurance, surety bonds, and commercial lines specifically exempted by Commissioner rule. |
CDOI Regulation 10-1-1 (3 CCR 702-10): Governance and Risk Management Framework
To implement SB 21-169, the Colorado Division of Insurance (CDOI) adopted Regulation 10-1-1 (codified at 3 CCR 702-10), titled Governance and Risk Management Framework Requirements for Life Insurers’ Use of External Consumer Data and Information Sources, Algorithms, and Predictive Models, which became effective on November 14, 2023 (with full implementation filings due December 1, 2024). While Regulation 10-1-1 was promulgated first for life insurers, CDOI has used its structural template as the baseline for subsequent private passenger automobile and health benefit plan stakeholder drafts.
Compared with the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers—which interprets existing Unfair Trade Practices Acts and Market Conduct Examination authority—Colorado Regulation 10-1-1 is a formal administrative regulation with mandatory components, board reporting duties, and officer attestation deadlines. For a side-by-side jurisdictional breakdown, see Insurer AI Governance: NAIC Model Bulletin vs. Colorado SB 21-169 vs. NY DFS Circular Letter No. 7.
| 3 CCR 702-10 Section | Governance Component | Mandatory Insurer Deliverables |
|---|---|---|
| Section 5.A.1–2 | Governing Principles & Board Oversight | Documented statement of values and principles ensuring ECDIS and predictive models do not result in unfair discrimination; clear lines of accountability up to the Board of Directors or a designated board committee. |
| Section 5.A.3 | Cross-Functional Governance Committee | Formal committee composed of representatives from actuarial, data science, underwriting, claims, legal, compliance, and consumer-advocacy/risk functions meeting at regular intervals. |
| Section 5.A.4–5 | Comprehensive ECDIS & Model Inventory | Up-to-date inventory of every ECDIS, algorithm, and predictive model deployed in production, including version history, model purpose, input variables, outputs, and material change triggers. |
| Section 5.A.6–7 | Risk Prioritization & Lifecycle Controls | Documented rubrics for prioritizing models by consumer impact, selecting training/validation datasets, testing for unfair discrimination prior to deployment, and monitoring ongoing model drift. |
| Section 5.A.8 | Third-Party Vendor Oversight | Insurers remain 100% legally responsible for ECDIS, scores, and algorithms supplied by external vendors; contracts must establish audit rights and documentation access. |
| Section 6 | Reporting & Negative Attestation | Insurers using ECDIS must submit annual narrative reports summarizing governance compliance. Insurers that do not use any ECDIS or ECDIS-driven models must file an annual officer-signed negative attestation. |
How Is Colorado SB 21-169 Used in Clinical Research, Health Insurance, and Biomedical Informatics?
In healthcare informatics, actuarial research, and AI in health insurance, SB 21-169 and its CDOI rulemaking proceedings directly govern how health benefit plans, disability carriers, and long-term care insurers ingest clinical and non-clinical datasets:
- Social Determinants of Health (SDOH) vs. ECDIS Boundaries: Health plans increasingly ingest third-party SDOH data (neighborhood deprivation indices, food-access metrics, transportation data, housing stability) to stratify patients for care management outreach. Under CDOI’s health benefit plan stakeholder framework, insurers and health informatics teams must distinguish between using SDOH/ECDIS solely to allocate supportive clinical care management resources versus using those variables in rating, underwriting, benefit design, or utilization management in ways that disadvantage protected populations.
- Prescription History, Medical Credit, and Wearable Telemetry in Accelerated Underwriting: Life and disability insurers frequently replace invasive paramedical blood/urine exams with accelerated underwriting models combining prescription drug databases (Rx histories), medical information bureau (MIB) codes, electronic health record (EHR) summaries, and consumer credit attributes. Under SB 21-169, when third-party consumer attributes or lifestyle proxies are blended with clinical data inside a gradient-boosted tree or neural risk score, the entire predictive model is subject to 3 CCR 702-10 governance and quantitative disparate-impact testing.
- Race/Ethnicity Imputation (BIFSG) and Quantitative Bias Testing: Because insurers generally do not collect race or ethnicity on insurance applications (and are often prohibited from doing so at point of sale), CDOI’s quantitative testing rules require insurers to estimate protected-class demographics using validated statistical imputation—specifically Bayesian Improved First Name Surname Geocoding (BIFSG)—and then run regression and approval-rate disparity tests to verify that ECDIS variables are not acting as statistical proxies for race or ethnicity after controlling for legitimate mortality or morbidity risk.
Common Misunderstandings About Colorado SB 21-169
- Misunderstanding 1: “SB 21-169 is the same law as the Colorado AI Act (SB 24-205).”Correction: SB 21-169 (enacted in 2021, codified at C.R.S. § 10-3-1104.9) is an insurance-specific statute enforced exclusively by the Colorado Division of Insurance. The later Colorado AI Act (SB 24-205, enacted in 2024) governs high-risk AI systems across general commercial sectors (employment, housing, lending, healthcare services) and explicitly exempts insurers that are already subject to C.R.S. § 10-3-1104.9 and CDOI’s rules.
- Misunderstanding 2: “Using a third-party vendor’s black-box score shields an insurer from SB 21-169 liability.”Correction: Under C.R.S. § 10-3-1104.9 and Section 5.A.8 of Regulation 10-1-1, an insurer cannot delegate regulatory compliance to a data broker or analytics vendor. If a third-party vendor refuses to disclose the variables or testing documentation needed to demonstrate compliance, the insurer cannot lawfully deploy that vendor’s score for Colorado policyholders.
- Misunderstanding 3: “Insurers that do not use ECDIS have no filing obligations under 3 CCR 702-10.”Correction: Even if a covered insurer uses zero ECDIS and zero predictive models incorporating ECDIS, Section 6 of 3 CCR 702-10 requires the carrier to submit a formal, officer-signed annual attestation to the Colorado Division of Insurance confirming non-use.
Frequently Asked Questions
What is colorado sb 21-169?
Colorado SB 21-169 (codified at C.R.S. § 10-3-1104.9) is a 2021 Colorado law that prohibits insurance companies from using External Consumer Data and Information Sources (ECDIS), algorithms, and predictive models in a way that unfairly discriminates against consumers based on race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression. It requires the Colorado Division of Insurance to adopt line-specific rules governing insurer AI governance and quantitative bias testing.
How is colorado sb 21-169 used in clinical research and healthcare?
In healthcare and clinical informatics, SB 21-169 governs how health benefit plans, disability insurers, and life insurance accelerated-underwriting systems combine clinical records (such as prescription drug histories and EHR summaries) with non-traditional consumer data (such as socioeconomic indicators, consumer credit attributes, or wearable activity streams). It also provides an operational methodology—using Bayesian Improved First Name Surname Geocoding (BIFSG) and covariate-adjusted regression—for auditing healthcare and actuarial predictive models for racial and ethnic proxy bias.
Sources
- Colorado General Assembly. Senate Bill 21-169: Restricting Insurers’ Use of External Consumer Data and Information Sources (Codified at C.R.S. § 10-3-1104.9). Signed July 6, 2021. https://leg.colorado.gov/bills/sb21-169
- Colorado Division of Insurance (CDOI). Regulation 10-1-1 (3 CCR 702-10): Governance and Risk Management Framework Requirements for Life Insurers’ Use of External Consumer Data and Information Sources, Algorithms, and Predictive Models. Effective November 14, 2023. https://doi.colorado.gov/for-consumers/sb21-169-protecting-consumers-from-unfair-discrimination-in-insurance-practices
- National Association of Insurance Commissioners (NAIC). Model Bulletin on the Use of Artificial Intelligence Systems by Insurers. Adopted December 4, 2023. https://content.naic.org/







