Overview
A consequential decision is a foundational concept in the Colorado Artificial Intelligence Act. It was introduced by the original 2024 law (SB 24-205) and carried forward when SB 26-189 (signed May 14, 2026) repealed and reenacted the statute, effective January 1, 2027. It defines when the law's consumer protection obligations are triggered.
Under the reenacted Colorado AI Act, a consequential decision is one that:
- Has a material, legal, or similarly significant effect on a consumer's life, and
- Concerns the consumer's access to, or the cost of, one of seven protected domains
If automated decision-making technology (ADMT) makes, or materially influences, a consequential decision affecting a Colorado consumer, and the deployer does business in Colorado, the deployment is subject to the Act's disclosure and consumer-rights obligations.
Protected Domains
The reenacted Colorado AI Act defines consequential decisions as those affecting a consumer's access to or the cost of:
| Domain | Examples |
|---|---|
| Employment | Hiring, termination, compensation, promotion, performance reviews |
| Education | Admissions, financial aid, academic assessments |
| Housing | Rental applications, mortgage approvals, real estate transactions |
| Financial or lending services | Credit approval, loan terms |
| Insurance | Eligibility, premium pricing, claims decisions |
| Health care | Diagnosis, treatment recommendations, coverage decisions |
| Essential government services | Benefits eligibility, licensing |
(The repealed 2024 law also listed legal services; that domain was not carried forward. The statute likewise excludes several categories from "consequential decision": advertising, marketing, product recommendations, search, and content moderation; cybersecurity, anti-fraud, anti-money-laundering, and sanctions-compliance tools; tools that only summarize or organize information for human review; and low-stakes routine processes.)
"Substantial Factor" Standard
"Substantial factor" was the standard in the repealed 2024 law (SB 24-205) and still appears in older guidance, but it is no longer the law. SB 26-189 replaced it with "materially influences," a defined term (C.R.S. 6-1-1701(13)): the ADMT's output must affect the outcome of the decision, for example by constraining it, or by ranking, scoring, recommending, or classifying. Incidental, trivial, or clerical uses do not qualify. Like its predecessor, the standard prevents employers, lenders, and others from using AI as a screen but claiming a human "made the decision."
ADMT likely materially influences a decision when:
- Its output is the primary basis for a decision (even if a human signs off)
- A human routinely follows the AI's recommendation without independent analysis
- The AI flags candidates for human review and humans only review flagged candidates
Comparison: Colorado vs. EU AI Act
Both the Colorado AI Act and the EU AI Act center their high-risk frameworks on decisions that significantly affect individuals in sensitive domains. The key differences:
| Colorado AI Act | EU AI Act | |
|---|---|---|
| Key term | Consequential decision | High-risk AI use case (Annex III) |
| Standard | Materially influences (defined term) | Substantial factor or intended use |
| Domains | 7 defined categories | 8 Annex III categories (broadly similar) |
| Consumer right | Data correction + human review | Human oversight mechanism |
Practical Classification Exercise
To determine if your AI system makes consequential decisions under Colorado law:
- What outputs does the system produce? (score, ranking, recommendation, classification)
- Do those outputs affect one of the seven listed domains? (And do none of the statutory exclusions apply?)
- Are Colorado consumers affected, and do you do business in Colorado?
- Does the AI materially influence the final decision? (Consider: what would happen if the AI were removed from the process?)
If all four are yes, you are likely deploying covered automated decision-making technology under the Colorado AI Act, with obligations effective January 1, 2027.