Colorado AI Law Deployer Checklist (SB 26-189)
A deployer under SB 26-189 is a person doing business in Colorado that deploys a covered ADMT. The checklist below walks through scope, notices, the 30-day post-adverse-outcome disclosure, consumer rights, and three-year records, flagging which items come from the statute and which only from the AG's proposed draft.
Last reviewed: November 13, 2026 · Independent coverage by Synthetic Data News · Editorial standards
Not legal advice. This article is general information from an independent publication, based on the public sources listed below. It is not legal advice and does not create an attorney-client relationship. Consult qualified counsel about your situation.
How to use this checklist
Items marked (statute) come from the enacted text. Items marked (draft) come from the AG's October 6, 2026 interim draft, which is proposed and subject to change. Meeting these items does not guarantee compliance with the law; compliance with part 17 is also not a defense to other law (C.R.S. 6-1-1709(2)).
1. Scope
- Do we do business in Colorado and deploy technology that processes personal data and generates scores, rankings, recommendations or similar output (statute, 6-1-1701(2), (7))?
- Is any output a non-de minimis factor that affects the outcome of a decision in a covered domain, such as employment, lending, housing, insurance, education, health care or essential government services (statute, 6-1-1701(3), (6), (13))?
- Have we checked the statutory exclusions for ADMT and for consequential decisions, such as tools used solely to summarize or route information for human review and low-stakes decisions (statute, 6-1-1701(2)(b), (3)(b))?
- Have we checked sector provisions: insurers subject to 10-3-1104.9, HIPAA covered entities, FDA-overseen devices, GLBA, FERPA and credit-notice rules (statute, 6-1-1704(6), (9), 6-1-1708)?
- Have we applied the draft presumption: output that sets a threshold, constrains options or produces a score about the individual, is reviewed by the decision-maker, and matches the outcome is presumed to materially influence, rebuttable by evidence it was a de minimis factor (draft Rule 6.1)?
2. Vendors and developers
- For each vendor system, have we requested the developer documentation: intended and harmful uses, training-data categories, limitations, monitoring and human-review instructions, and what we need for our notices (statute, 6-1-1702(1))?
- Do we have a way to learn about material updates and changes to intended use (statute, 6-1-1702(2))?
- If a third party runs the system for us, have we recorded that we remain responsible for deployer obligations and reflected duties to assist in the contract (draft Rule 4.3)?
- Have we reviewed contract indemnity terms against the rule that provisions indemnifying a party for its own discrimination-law violations are void (statute, 6-1-1707(7))? Ask counsel.
3. Notice before use
- Before using a covered ADMT to materially influence a decision, do we give a clear and conspicuous notice that we use or will use it, with instructions for getting more information (statute, 6-1-1704(1))?
- If relying on a public posting, is it prominent and reasonably accessible at points of consumer interaction, such as a link near the transaction (statute, 6-1-1704(2))?
- Are notices accessible to consumers with disabilities and limited English proficiency (statute, 6-1-1704(8); draft Rule 3.2)?
4. Adverse outcome disclosure within 30 days
- Can we identify adverse outcomes: denials, terminations, revocations, material reductions, or materially less favorable pricing or terms (statute, 6-1-1701(1))?
- Within 30 days after the decision, do we provide a plain-language description of the decision and the ADMT's role (statute, 6-1-1704(3)(a))?
- Do we explain how to request more information, including the ADMT's name, version number if applicable, developer, and types, categories and sources of personal data used, to the extent we receive that information from the developer (statute, 6-1-1704(3)(b))?
- Do we explain consumer rights under 6-1-1705 and how to exercise them (statute, 6-1-1704(3)(c))?
- If withholding trade-secret or legally protected information, do we notify the consumer (statute, 6-1-1704(5))?
- If we are a creditor sending federal ECOA or FCRA notices, have we checked whether the federal notice also meets part 17 and may carry a brief ADMT statement (statute, 6-1-1704(6))?
- Have we reviewed the draft's rules on form (written, including electronic) and content (draft Rules 7.2 to 7.4) and will we re-check the final rule before January 1, 2027?
5. Consumer rights
- After an adverse outcome, can a consumer request instructions for obtaining and correcting factually incorrect or materially inaccurate personal data used in the decision (statute, 6-1-1705(1)(a)(I))?
- Do we know correction is not required for opinions, predictions, scores or protected evaluations (statute, 6-1-1705(1)(c))?
- Can a consumer get meaningful human review and reconsideration, to the extent commercially reasonable (statute, 6-1-1705(1)(a)(II))?
- Does our reviewer meet the statutory definition: designated by us, with authority to approve, modify or override; considers available primary evidence; trained; does not default to the system output; has enough information about intended use, limitations, input categories and principal factors (statute, 6-1-1701(15))?
- FERPA-subject education deployers: have we mapped these rights to existing student-record procedures (statute, 6-1-1705(2))?
- Have we reviewed the draft's reviewer standards, such as authority to approve, modify or override, training, and the proposal that ADMT may not assist the review (draft Rule 8.7)?
6. Records
- Do we retain, for at least three years after the decision, records reasonably necessary to demonstrate compliance, which may include version identifiers, changelogs and documentation of material mitigation changes (statute, 6-1-1703)?
- If conducting human reviews, do we preserve contemporaneous documentation such as reviewer identity and authority, evidence considered and the outcome (draft Rule 8.7(H), proposed)?
7. Enforcement and monitoring
- Do we understand enforcement is by the Attorney General under the Colorado Consumer Protection Act and there is no new private right of action (statute, 6-1-1706(1), (4), 6-1-1709(1))?
- Do we understand the AG issues a notice of violation if it deems a cure possible, with 60 days to cure, and that no cure period is required for knowing or repeated violations (statute, 6-1-1706(3))? That subsection is repealed effective January 1, 2030.
- Is someone assigned to watch for the adopted rule (eDocket 2026-00360) and to update this checklist?
Frequently asked questions
Who counts as a deployer in Colorado?
A person doing business in Colorado that deploys a covered ADMT (C.R.S. 6-1-1701(7)).
How long do deployers have to send an adverse outcome disclosure?
Within thirty days after making the decision (C.R.S. 6-1-1704(3)).
How long must deployers keep records?
At least three years after the date of a consequential decision, or longer if other law requires (C.R.S. 6-1-1703).
If a vendor runs the AI for us, are we still responsible?
The AG's proposed draft says the deployer remains responsible for deployer obligations where a third party runs the system at its direction and it makes the decision (draft Rule 4.3). This is proposed.
Sources
- Enacted SB 26-189, Session Laws of Colorado 2026, ch. 131 (C.R.S. 6-1-1701 to 6-1-1709), approved May 14, 2026
- Colorado Department of Law, ADMT & Chatbot rules, interim draft redline, October 6, 2026 (proposed; subject to change)
- Colorado Department of Law, Regulatory Analysis, October 21, 2026
Facts in this article were checked against these sources on November 13, 2026. Draft rule content is proposed and subject to change.
Other resources
- Colorado Attorney General: AI rulemaking — Official rulemaking page.
- Colorado eDocket 2026-00360 — Official rulemaking record.
- CertifiedData.io: Colorado SB 26-189 guide (vendor resource) — A vendor resource from a platform operated by the same organization as SDN (see our editorial standards). One of several options for organizing evidence, alongside internal logging, general governance and model-risk tooling, and open-source tooling. SDN has not evaluated its claims.
Get the SDN weekly digest
Independent AI governance and synthetic data coverage, once a week.
Related coverage
Not legal advice. This article is general information from an independent publication, based on the public sources listed below. It is not legal advice and does not create an attorney-client relationship. Consult qualified counsel about your situation.