Rights & Ethics

Unfair Discrimination

Practices where AI systems lead to biased or inequitable treatment of consumers without legitimate grounds.

How the laws define this

Regulators agree that unfair discrimination involves treating individuals or groups unequally, differently, or adversely within insurance practices, particularly when utilizing predictive models, algorithms, or artificial intelligence systems. The key point of divergence is how instruments evaluate whether differential treatment is permissible. Under Colorado SB 21-169 - AI in Insurance Underwriting, discrimination is defined directly by prohibited status, broadly banning practices based on listed protected characteristics such as race, sex, disability, or gender identity. Conversely, Vermont Act 89 - AI in Insurance and the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers condition unfairness on justification, defining it as differential treatment that lacks a legitimate, non-discriminatory, actuarial, or underwriting basis. Taking an outcome-focused approach, CONCERNING PROTECTING CONSUMERS FROM UNFAIR DISCRIMINATION IN INSURANCE PRACTICES. targets external consumer data and predictive models that correlate with protected characteristics, defining unfair discrimination as disproportionately negative outcomes for these groups that exceed a reasonable correlation to underlying factors like losses and costs.

Synthesised from the 4 statutory definitions below. Each one is quoted in full, with its source.

How jurisdictions differ

What is distinctive in each definition, binding law in force first. Showing 3 of the 4 instruments that define this term; all are listed under Definitions below.

Jurisdiction	Instrument	Status	What is distinctive
Colorado, United States	Colorado SB 21-169 - AI in Insurance Underwriting	In Force (Amended)	Explicitly lists protected characteristics like race, sex, and gender identity, limiting discrimination to those categories in insurance practices unlike the others.
Vermont, United States	Vermont DFR Bulletin No. 229 — The Use of Artificial Intelligence Systems in Insurance	In Force	Emphasizes adverse outcomes for policyholders or applicants, focusing on decision-making impacts without listing specific protected traits.
United States	NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers	In Force	Specifies unequal treatment lacking a legitimate actuarial or underwriting basis, narrowing to insurance decisions without mentioning protected characteristics.

What it triggers

In Colorado, insurers are prohibited from using algorithms or predictive models that cause unfair discrimination.source
Colorado SB 21-169 requires risk frameworks and bias tests.source
Vermont DFR Bulletin No. 229 sets out guidance for insurers on deploying AI systems.source
The NAIC Model Bulletin guides insurers on managing governance, risk, and regulatory compliance when deploying AI.source

Definitions (4)

The practice of treating individuals or groups differently without a legitimate, non-discriminatory basis, particularly when AI systems are used in insurance to make decisions that result in adverse outcomes for policyholders or applicants.

Vermont Act 89 - AI in InsuranceDefinition 1 of 4

Discrimination based on an individual's race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression in any insurance practice.

Colorado SB 21-169 - AI in Insurance UnderwritingDefinition 2 of 4

The practice of treating individuals or groups unequally in insurance decisions without a legitimate actuarial or underwriting basis, often a key risk addressed by AI regulation.

NAIC Model Bulletin on the Use of Artificial Intelligence Systems by InsurersDefinition 3 of 4

This term refers to the use of external consumer data, algorithms, or predictive models that correlate with protected characteristics (e.g., race, sex) and result in disproportionately negative outcomes for these groups. The negative outcome must exceed a reasonable correlation to underlying insurance factors like losses and costs.

CONCERNING PROTECTING CONSUMERS FROM UNFAIR DISCRIMINATION IN INSURANCE PRACTICES.Definition 4 of 4

Not to be confused with

Unlawful Discrimination or Disparate Impact

Unlawful discrimination or disparate impact focuses on unjustified differential treatment based on protected characteristics, while unfair discrimination definitions here vary in specifying those traits or bases.

Harmful Bias

Harmful bias refers to systematic AI errors causing unfair outcomes, whereas unfair discrimination targets biased treatment in insurance practices without legitimate grounds.

Biased output

Biased output means unjustified adverse impacts on protected groups, differing from unfair discrimination by not always requiring insurance-specific contexts or actuarial bases.

Questions and answers

What protected classes does Colorado law cover under unfair discrimination?

Colorado SB 21-169 defines it based on race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression.

Does the NAIC bulletin prohibit AI use in insurance?

The NAIC Model Bulletin is a guideline that describes practices to manage risks like unfair discrimination.

What must Vermont insurers do about AI under Bulletin 229?

Vermont DFR Bulletin No. 229 guides insurers on AI deployment to avoid unfair discrimination resulting in adverse outcomes.