On August 11, 2026 the Colorado Department of Law filed proposed rules implementing the ADMT Act (SB 26-189) and the Chatbot Safety Act (HB 26-1263). The draft adds a 'midstream developer' category, offers two alternative tests for when ADMT 'materially influences' a decision, prescribes the contents of post-adverse-outcome disclosures (principal reasons with real specificity, whether inferences or risk scores drove the result, any automatic-denial factors), and sets standards for meaningful human review — independent, subject-matter-competent reviewers with override authority who may not themselves rely on ADMT. Comments to shape the revised draft are due September 4, 2026; a revised draft is targeted for September 23; the hearing and final comment deadline are October 26, 2026.
| Document | Colorado AG proposed ADMT rules — Proposed Automated Decision-Making Technology and Conversational AI Service Rules (Notice of Rulemaking Hearing) |
| Issued by | Colorado Artificial Intelligence Act (SB 24-205), repealed and reenacted as the Automated Decision-Making Technology Act (SB 26-189) — enforced by the Colorado Attorney General |
| Type | Regulation |
| Status | Comment period open |
| Published | Aug 11, 2026 |
| Comment deadline | Oct 26, 2026 |
| Applies to | Developers, 'midstream developers', and deployers of covered ADMT under SB 26-189, and operators of conversational AI services under HB 26-1263 |
| Official source | coag.gov ↗ |
| Use cases | Credit scoring & underwriting · Fair lending & discrimination · Third-party & vendor AI · Customer-facing chatbots · AI governance (general) |
What are the key points of Colorado AG proposed ADMT rules?
- Filed with the Colorado Secretary of State on August 11, 2026 as an expedited rulemaking to meet the January 1, 2027 statutory deadline
- Creates 'midstream developers' — firms that integrate covered ADMT into their own products — who must obtain upstream documentation and pass it downstream
- Two alternative 'materially influences' tests: whether the decision-maker reached an independent judgment after reviewing the output, or whether other information played a substantially larger role than the ADMT output; the AG notes the second is easier to satisfy
- Adverse-outcome disclosures must state the decision, the deployer's purpose, human reviewer roles, effective dates, the principal reasons with specificity, and whether inferences or risk scores drove the outcome — generic references to internal policy are insufficient, and disclosures must be tailored to the sector (e.g. lending)
- Meaningful human review: reviewers independent from the original decision-maker where feasible, trained, insulated from managerial pressure, empowered to override, and barred from using ADMT to conduct the review; 'commercially reasonable' scales with harm severity and deployer capacity
- Also implements HB 26-1263 disclosure, age-estimation, and reporting duties for conversational AI operators
- Deadlines: September 4, 2026 (comments for revised draft), September 23, 2026 (revised draft target), October 26, 2026 (hearing in Denver/videoconference and final written comments)
What did Colorado AG proposed ADMT rules change for banks?
This is where the ADMT Act's bank-relevant detail is being decided. The 'materially influences' test determines whether a scorecard that feeds an underwriter's judgment is covered at all; the disclosure rules go beyond the reason-code conventions of Reg B adverse action notices; and the human-review standard would require most lenders to formalize an appeal channel staffed independently of the original credit decision. Banks with Colorado customers have until October 26, 2026 to comment.
When are comments due on Colorado's ADMT rules?
September 4, 2026 to be considered for the revised draft, and October 26, 2026 (11:59 PM MST) as the final deadline coinciding with the public hearing.
Does the AG's human-review standard let a bank use AI to review an appeal?
Not as drafted. The proposed rules say reviewers may not use ADMT to conduct the review and should be independent, trained, and able to override the outcome.
| Date | Document | Status |
|---|---|---|
| May 29, 2026 | HB 26-1263 — Conversational Artificial Intelligence Service Operator Requirements (Chatbot Safety Act) | Final |
| May 14, 2026 | SB 26-189 — Automated Decision-Making Technology Act (repeal and reenactment of the Colorado AI Act) | Final |
| Aug 28, 2025 | SB 25B-004 — Artificial Intelligence Consumer Protections — delay of effective date to June 30, 2026 | Superseded |
| May 17, 2024 | SB 24-205 — Consumer Protections for Artificial Intelligence (Colorado Artificial Intelligence Act) | Superseded |
| Jul 24, 2026 | Regulation (EU) 2026/1744 (Digital Omnibus on AI) — Regulation (EU) 2026/1744 amending Regulation (EU) 2024/1689 as regards the simplification of the implementation of harmonised rules on artificial intelligence (Digital Omnibus on AI) | In force |
| May 19, 2026 | Draft Commission guidelines on high-risk classification — Draft Commission Guidelines on the classification of high-risk AI systems under Article 6 of the AI Act | Proposed |
Follow every document these regulators publish
6 curated AI stories for banking executives · Every morning · Free
Subscribe to BankingNewsAI →