Skip to main content

Colorado AI Law Explained From the Statute (SB 24-205 and SB 26-189)

Title card for the Vaquill AI guide: Colorado AI Law Explained From the Statute (SB 24-205 and SB 26-189)

Short answer: the Colorado AI law, Senate Bill (SB) 24-205, never took effect. In May 2026 lawmakers repealed it and passed a new law, Senate Bill 26-189. The new law covers software that helps decide who gets a job, a loan, a home, insurance, care, a school place or a public benefit. It starts January 1, 2027. Only the state attorney general (the AG) can enforce it, and the rules are still being written. A federal court order the AG agreed to also holds back AG enforcement for now, on our reading of its wording. Skip the old impact-assessment program. Build notices, 30-day letters and vendor document requests. This is not legal advice. Ask a lawyer whether it applies to you.

Reviewed and corrected on 12 October 2026. This is general information, not legal advice; check the cited statute or opinion before relying on it.

Official text, as of 12 October 2026

Colorado Revised Statutes § 6-1-1706(4), as reenacted by SB 26-189 (Laws of 2026, chapter 131, approved May 14, 2026): "Nothing in this part 17 creates a new private right of action." Section 6-1-1704(3) requires a deployer to give an affected consumer a plain language description of the decision and the technology's role "within thirty days after making the decision." Read the signed act on the General Assembly's bill page or as the session law PDF.

What changed in 2026: SB 26-189 repealed and reenacted part 17, dropping the 2024 duty of reasonable care, risk management program and impact assessments for a notice, explanation and review model. Most of it takes effect January 1, 2027.

TL;DR

  • Status on October 12, 2026: the old Act was replaced before it applied to anyone. New duties start January 1, 2027.
  • Who: developers who build covered tools and deployers (businesses that use them) doing business in Colorado.
  • What: developers share documents. Deployers give notice, explain bad outcomes within 30 days, and offer fixes and human review.
  • Enforcement: the state AG only. The statute creates no new private right of action, meaning no new right for individuals to sue under it.
  • Still moving: draft rules are open for comment through October 26, 2026 (longer if the hearing runs past that day), and may change. In a federal case, the AG agreed to a court order that, on our reading of its wording, bars it from starting enforcement for now, and the case deadlines are stayed. Re-check before you finalize notices.

Why the old Colorado AI law name still confuses people

Most guides still describe SB 24-205, signed in 2024 with duties set for February 1, 2026. A special session bill moved that to June 30, 2026. Then SB 26-189 arrived. The AG's own page says it "repeals and reenacts" the 2024 provisions.

Both laws sit in one place in the Colorado Revised Statutes (C.R.S.), the state's official set of laws: Title 6, article 1, part 17, from section 6-1-1701. Repeal and re-enact means the old text came out and new text took its slot. The numbers look the same, but the content changed. For how bills land there, see what a session law is.

The dates, in order

  • May 17, 2024: SB 24-205 signed. Duties were set for February 1, 2026.
  • August 28, 2025: SB 25B-004 signed. New start date: June 30, 2026.
  • April 9, 2026: X.AI LLC sued the AG in federal court in Colorado (X.AI LLC v. Weiser, No. 1:26-cv-01515). On April 24 the United States filed a complaint as plaintiff-intervenor.
  • April 27, 2026: a magistrate judge granted the parties' joint motion and stipulation and ordered that "the defendant shall not initiate enforcement, including but not limited to the initiation of an investigation, for alleged violations of SB24-205 (or any legislation replacing or amending SB24-205 enacted during this legislative session)." All case deadlines were stayed too (the order).
  • What that means: the bar covers violations "that occurred or may occur on or before 14 days after the date the Court issues a ruling" on X.AI's "forthcoming" preliminary injunction motion (a request for a court order to stop enforcement). X.AI must file that motion within 28 days after final adoption of the AG's rules.
  • The catch: on its face, the wording reaches SB 26-189, which was enacted in the same session and replaces SB 24-205. That is our reading. As of the last docket entry we saw (October 2, 2026, a lawyer's withdrawal), we found no injunction motion or ruling. Check the docket before you plan around it.
  • May 14, 2026: SB 26-189 signed (Laws of 2026, chapter 131). Three provisions took effect on passage: the AG's duty to adopt rules on the post-adverse-outcome disclosures (section 6-1-1704(4)), the same duty for the consumer rights (section 6-1-1705(3)), and the joinder clause (section 6-1-1706(6)).
  • July 1, 2026: the Federal Trade Commission proposed a policy statement, voted 2-0, saying Colorado's Artificial Intelligence Act is "impliedly preempted to the extent it conflicts with a federal regulatory scheme." Comments closed July 31, 2026, and we found no final version. An earlier executive order, EO 14365 (90 Fed. Reg. 58499), also names Colorado's law. Neither is a statute that overrides it.
  • August 11, 2026: the AG filed proposed rules. An interim draft followed on October 6. A hearing is set for October 26, 2026 at 10:00 a.m. in Denver and online. Written comments are open through October 26 (or the last hearing day, if it runs longer).
  • January 1, 2027: new duties start. The act "applies to consequential decisions made on or after January 1, 2027." The AG must adopt rules on the post-adverse-outcome disclosures and the consumer rights by that date.

What the law is

Say a business in Colorado uses software that scores or ranks people, and the output shapes a decision in one of seven listed areas. The business must tell the person. It must explain a bad outcome within 30 days. It must give instructions for asking to correct wrong data, and a chance at human review "to the extent commercially reasonable." The company that built the software must give its customers the right documents. The 2024 version also required impact assessments, which are written risk reviews. The 2026 version dropped them.

Made-up example: a Denver employer ranks resumes with a tool. It tells applicants first. If one is rejected, the employer explains within 30 days. The applicant can ask how to correct wrong data and ask for a human review.

Who it applies to

The statute calls the technology "automated decision-making technology," or ADMT. In plain terms, it is software that scores or ranks people. Under section 6-1-1701(2)(a), it is technology that "processes personal data and uses computation to generate output," like predictions, rankings and scores. That output must be "used to make, guide, or assist a decision."

The law bites only when the tool is "used to materially influence a consequential decision." That is section 6-1-1701(5). Section 6-1-1701(13) says an output does this when it "affects the outcome" and is "a non-de minimis factor," meaning more than a trivial one. A consequential decision is about a person's access to a "covered domain." Section 6-1-1701(6) lists seven. They are education, employment, buying or leasing residential real estate in Colorado, financial services, insurance, health care, and essential government services. The definition also reaches a decision about price or other material terms that is reasonably likely to materially limit, delay or deny access to one of those domains (section 6-1-1701(3)(a)(II)). Incidental, trivial or clerical uses do not count as materially influencing a decision (section 6-1-1701(13)(b)).

A developer builds, sells or licenses a covered tool, or builds a part made to go inside one. A deployer uses one. Both must do business in Colorado. A "consumer" includes an employee, a job applicant who is a Colorado resident, and anyone whose access or eligibility in Colorado is judged in a consequential decision by a person doing business in Colorado (section 6-1-1701(4)(b)). We found no company-size threshold in the text.

The 2024 law used other words. It covered a "high-risk artificial intelligence system," meaning one that "makes, or is a substantial factor in making, a consequential decision." That phrase is gone now.

What each side must do

DutyWho it applies toWhere in the statute
Give deployers a statement of intended uses, harmful uses, training data types and limitsDeveloper§ 6-1-1702(1)
Tell deployers about material updates and changes in intended useDeveloper§ 6-1-1702(2)
Keep compliance records at least three yearsDeveloper and deployer§§ 6-1-1702(4), 6-1-1703
Give clear notice before using the tool in a decision (a prominent public notice near the point of interaction can satisfy this)Deployer§ 6-1-1704(1), (2)
Within 30 days of a bad outcome, describe the decision and the tool's role, and explain how to ask for more information and use the consumer rightsDeployer§ 6-1-1704(3)
On request after a bad outcome, give instructions for requesting personal data and correcting factually incorrect data, and offer human review and reconsideration "to the extent commercially reasonable"Deployer§ 6-1-1705(1)
Make notices usable for people with disabilities or limited EnglishDeveloper and deployer§ 6-1-1704(8)

Human review has a set meaning in section 6-1-1701(15). The reviewer must have power to "approve, modify, or override" the decision. The reviewer must consider relevant primary evidence, be trained, and must not "default to the system output." Section 6-1-1705(1)(c) says the correction right does not reach opinions, predictions, scores or protected evaluations.

What counts as discrimination now

The 2024 text defined "algorithmic discrimination" and required "reasonable care" to avoid it. The 2026 act never uses that phrase. It leans on laws already in place. Section 6-1-1707(1) says a developer or deployer "may be held liable" in a discrimination claim under the state Anti-Discrimination Act when a covered tool materially influenced the decision. Blame is split by "relative fault," meaning each side answers for its own share, and a developer is liable only to the extent the tool was used in a way the developer intended, documented or marketed (subsection (5)). A contract clause that indemnifies a party against liability for its own acts or omissions in violation of Colorado anti-discrimination law is "contrary to public policy and void," meaning a court will not enforce it. Other commercial terms between businesses stay enforceable (subsection (7)(c)). And using the tool "does not excuse, justify, or provide a defense to any obligation or liability" under state or federal law (subsection (9)).

What it does not cover

  • Basic tools such as calculators, spell-check, firewalls, databases, and spreadsheets that need human analysis and do not use machine learning, foundation models or large language models. Source: section 6-1-1701(2)(b)(I).
  • A tool used by an individual "solely to summarize, organize, translate, draft, route, or present information for human review of administrative processing" (section 6-1-1701(2)(b)(II)).
  • Advertising, search, content moderation, scheduling and customer service triage. Source: section 6-1-1701(3)(b).
  • Chat tools not marketed or set up for consequential decisions, if a written use policy (rules on how customers may use the tool) bars that use.
  • Covered entities under the federal health privacy law (HIPAA) and their business associates, except for hiring decisions. A health care provider qualifies only if it operates from a location in Colorado, and covered entities must still give patients a general notice of advanced technology use and certain disclosures about financial assistance decisions (section 6-1-1708(3)).
  • Medical devices overseen by the Food and Drug Administration, and a drug or device maker's FDA-overseen research (section 6-1-1708(4)).
  • Insurers subject to the state insurance algorithm rules in section 10-3-1104.9, which "are in compliance" in the practice of insurance. Employment uses by insurers stay covered (section 6-1-1708(1), (2)).
  • Creditors that already give the federal credit notices, and schools with federal student-record procedures, can use those notices (sections 6-1-1704(6) and (9), 6-1-1705(2)).
  • Trade secrets. A business that withholds information must tell the consumer.

Who enforces it

The AG enforces the law under the state Consumer Protection Act. Under section 6-1-1706, a violation is a "deceptive trade practice," which is conduct the state treats as unfair or misleading to consumers. If the AG deems a cure possible, it sends a notice first, and the business has 60 days to fix the problem. A knowing or repeated violation gets no cure period. A court may treat a cure within 60 days of written notice as a mitigating factor when it sets penalties (section 6-1-1706(3)(d)). The cure-period subsection is repealed effective January 1, 2030.

On private lawsuits the text is direct: "Nothing in this part 17 creates a new private right of action." That is section 6-1-1706(4). Rights people already have, such as under the Anti-Discrimination Act, stay. Part 17 lists no penalty amounts. Look to the Consumer Protection Act.

What a business should check

  1. List every tool that scores or ranks people in hiring, lending, housing, insurance, health care, school or benefits decisions.
  2. For each, ask if its output really affects the outcome.
  3. Ask each vendor for the documents section 6-1-1702 lists. Put a due date in the contract.
  4. Draft the notice and the 30-day letter. Name a person who can override a decision.
  5. Check contracts that shift blame. Section 6-1-1707 voids some of them.
  6. Read the final AG rules once adopted, and watch the docket. Terms like "materially influence" are still being pinned down.

What goes wrong

The top mistake is building impact assessments from a pre-May 2026 article. The law no longer asks for them. The next is the reverse: thinking nothing applies. Notices and 30-day letters are real work. Last, do not take a vendor's word that it is "compliant." Ask for the documents.

This guide is part of US Law Data: The Complete Guide, a map of where US law comes from and how to use it.

Judgment call
Question 1 of 4

Your HR team bought a resume-ranking tool. A 2025 blog post told you to run a Colorado impact assessment and risk program. Where should the effort go now?

FAQ

Is the Colorado AI Act in effect? No. SB 24-205 was repealed and replaced before its duties applied. The new law, SB 26-189, starts January 1, 2027, except that the two provisions requiring the AG to adopt rules on the disclosures and consumer rights, and the joinder clause, took effect on passage.

What is the Colorado AI Act effective date? January 1, 2027, for decisions made on or after that day. Older articles list February 1, 2026 or June 30, 2026. Those dates belong to the repealed law.

Does my business have to be in Colorado? The duties fall on a "person doing business in Colorado." The statute also counts a job applicant who lives there as a consumer. Ask your lawyer if your business qualifies.

Do I need an impact assessment? Not under the current text. The 2024 law required one. The 2026 law does not.

Can someone sue me directly under it? The statute says it creates no new private right of action. Claims people already have, such as bias claims, remain.

Is a general chatbot covered? Only if it helps decide something in one of the seven areas. House Bill 26-1263 (Chapter 208, Laws of 2026) is a separate law for operators of conversational AI services. It adds definitions to the same part 17 and sets duties on chatbot operators. Most of them apply "on and after January 1, 2027" and the annual report to the AG starts July 1, 2027.

Where do I read the official text? The General Assembly's page for SB 26-189 has the signed act, and the session law PDF shows it in full. The AG's page tracks the rules, and SB 24-205 shows the original. To cite Colorado statutes correctly, see how to find and cite Colorado law. For how statutes and the rules built on them differ, see what a regulation is.

For the US statutes, regulations and court rules these laws sit in, see the US primary law API.

Connect our US primary law database.
Every US statute, regulation, constitution, and executive order via REST, MCP or SQL. 6M+ sections, section-level citations, and links to the official source. Plus a free open dataset.
Updated October 12, 202617 min read

New legal AI guides, weekly.

Priyansh Khodiyar

Priyansh Khodiyar

Co-Founder & CTO

Priyansh leads engineering and AI at Vaquill AI: the pipelines that pull statutes, regulations and court rules from every US jurisdiction's official publisher, and the REST API, MCP server and open dataset that serve them.

Cookie preferences

Choose what Vaquill may use on this browser. You can change this at any time from “Cookie settings” in the footer.

Details in the Cookie Policy.