Blog

Colorado AI Act 2027: What Small Businesses Must Do

Colorado replaced its 2024 AI law in 2026 and added a chatbot law. Both start on 1 January 2027. Here is what changes for small businesses that use AI tools.

Atiq Rezaul Hoque Turjo
Atiq Rezaul Hoque Turjo
10 min read
Colorado AI Act 2027: What Small Businesses Must Do

Your business might use AI every day without realizing that some of those tools could fall under Colorado's new AI rules.

The important distinction is not whether a tool uses artificial intelligence. It is what the tool does and whether its output influences an important decision about an individual.

The Colorado AI Act changes significantly on 1 January 2027. Colorado replaced its original AI law with an automated decision-making technology (ADMT) law and introduced separate chatbot safety requirements. Most routine customer-service and marketing tools are excluded, but AI used in hiring, lending, insurance and housing deserves closer attention.

Here is how small and mid-sized businesses can identify the tools that matter and prepare before the rules take effect.

What changed in the Colorado AI Act in 2026?

Colorado made two significant legislative changes in 2026.

First, Senate Bill 26-189 repealed and re-enacted the state's 2024 AI law, SB 24-205, as an automated decision-making technology (ADMT) law.

Second, House Bill 26-1263 introduced the Colorado Chatbot Safety Act, establishing separate requirements for certain conversational AI services.

Both laws take effect on 1 January 2027.

The revised ADMT law applies to consequential decisions made on or after that date. This means businesses should pay attention to how their AI systems operate when making decisions from the beginning of 2027, rather than treating the legislation as a requirement only for newly purchased software.

The Colorado Attorney General's office is responsible for developing rules that clarify how businesses must provide certain disclosures following adverse decisions.

The rulemaking timeline includes:

  • 11 August 2026: The Attorney General's office filed proposed ADMT and Chatbot Safety rules.
  • 6 October 2026: An interim draft was released.
  • 21 October 2026: Regulatory analysis was published.
  • 26 October 2026: A public hearing was held.
  • 17 November 2026: Final rules were not yet listed on the official rulemaking page when checked.

The Attorney General must adopt rules addressing post-adverse-outcome disclosures by 1 January 2027.

Businesses should monitor the Colorado Attorney General's AI rulemaking page for final requirements and updates.

The legislation itself is available through the official SB 26-189 bill page and HB 26-1263 bill page, with the enacted text in the SB 26-189 session law and the HB 26-1263 session law.

Does your AI tool count as automated decision-making technology?

The Colorado ADMT law does not automatically cover every application that uses AI. It focuses on technology that processes personal data and materially influences consequential decisions about individuals.

Automated decision-making technology (ADMT) means technology that uses computation to generate outputs such as:

  • Predictions
  • Recommendations
  • Classifications
  • Rankings
  • Scores

Those outputs must be used to make, guide or assist a decision about an individual.

The second part of the test is whether the decision is consequential.

Under SB 26-189, consequential decisions concern access to, eligibility for, selection for or compensation in areas including:

  • Employment: Hiring, selection and employment-related decisions.
  • Financial services: Lending and other financial services.
  • Insurance: Access to insurance and insurance-related eligibility.
  • Housing: Rental applications and housing opportunities.
  • Education: Access to educational opportunities.
  • Healthcare: Access to healthcare services.
  • Government services: Essential government services and public benefits.

The definition also includes differentiated prices or terms that materially limit someone's access to these opportunities or services.

However, the law contains important exclusions.

Routine business operations are generally outside the ADMT framework.

Examples include routine scheduling, administrative routing, customer-service triage, communicating decisions and workflow management.

Other excluded uses include:

  • Advertising and marketing.
  • Product recommendations.
  • Search and content moderation.
  • Tools that only summarize, organize or present information for human review without producing a score, ranking, recommendation or other inference that materially influences the outcome.

There is another important distinction: incidental, trivial or clerical uses do not count as materially influencing a decision.

For example, an AI assistant that organizes incoming job applications is different from a system that ranks candidates and determines which applications receive attention.

The first may simply support administration. The second may influence an employment decision.

Linen embroidery of a blank job application clipboard beside a magnifying glass, representing AI-assisted hiring decisions.

AI that ranks job applicants deserves closer scrutiny than software that simply organizes application information.

The practical lesson is simple: look at the role of the AI output, not just the name of the software.

What must businesses do when they deploy a covered AI tool?

If your business deploys covered ADMT, SB 26-189 introduces specific responsibilities.

The law distinguishes between developers who create ADMT systems and deployers who use them to make consequential decisions.

For most small businesses purchasing AI software, the deployer responsibilities are particularly important.

Provide notice before using covered AI

Businesses must provide consumers with clear and conspicuous notice before using ADMT to materially influence a consequential decision.

A prominent public notice at the points where consumers interact with the business can satisfy this requirement.

Explain adverse outcomes within 30 days

When a covered system contributes to an adverse consequential decision, the deployer must provide information within 30 days.

This includes:

  • A plain-language description of the decision.
  • An explanation of the role the ADMT played.
  • Information about how to request additional details.
  • An explanation of the consumer's applicable rights.

This is why businesses should not rely entirely on a software vendor to handle compliance. They need an internal process for identifying adverse outcomes and communicating with affected individuals.

Support correction and human review

Consumers have the right to request correction of factually incorrect personal data used in the decision.

They can also request meaningful human review and reconsideration.

That means a business needs an actual person who can examine the relevant information, consider the circumstances and reconsider the outcome.

A button that automatically sends a complaint back through the same AI system is not a meaningful human review process.

Embroidered desk lamp illuminating a folder beside two chairs, representing meaningful human review of AI decisions.

Human review should give people a genuine opportunity to have an AI-influenced decision reconsidered.

Keep records

Deployers must maintain relevant records for at least three years.

Businesses should establish a retention process covering decisions, notices, relevant system information and records needed to demonstrate how their procedures operate.

What about AI developers and enforcement?

Developers must provide deployers with technical documentation. This should cover intended uses, categories of data used, known limitations and risks, and instructions for appropriate use, monitoring and human review.

Enforcement belongs to the Colorado Attorney General under the Colorado Consumer Protection Act.

Before 1 January 2030, the Attorney General must provide 60 days' notice and an opportunity to cure where a cure is possible. The law does not create a new private right of action.

For a broader operational approach, see our AI agent governance framework and guidance on human approval gates for AI agents.

Need a structured approach? Explore AI governance for business to understand how to establish ownership, oversight and review processes for AI systems.

Does the Colorado Chatbot Safety Act apply to business chatbots?

Not necessarily. This is one of the most important distinctions for businesses using AI assistants.

HB 26-1263, the Colorado Chatbot Safety Act, covers certain publicly accessible conversational AI services primarily designed to simulate human conversation.

However, the law specifically excludes several categories of services.

These include services primarily designed to support commerce or transactional activities, such as:

  • Product recommendations.
  • Ordering and payments.
  • Delivery and returns.
  • Customer support and customer service.

It also excludes services designed and marketed for business operations or productivity.

This means a typical business website support chatbot, appointment-booking assistant or customer-service agent is likely excluded, provided it falls within the statutory exclusions.

That is different from a publicly accessible companion-style AI application primarily designed to simulate human relationships or conversation.

What happens when a chatbot is covered?

Operators of covered services must disclose that users are interacting with AI.

The disclosure must occur:

  • At the beginning of the first interaction each day.
  • At least every three hours during a continuous conversation, or through a persistent notice.
  • Whenever the user asks whether they are interacting with AI.

The law also introduces additional responsibilities.

Operators must maintain a suicide and self-harm response protocol, submit annual reports to the Attorney General and follow additional requirements concerning known minor users.

Covered services must not present their outputs as equivalent to services provided by licensed professionals.

For businesses building customer-facing assistants, this is another reason to define what the system can and cannot represent itself as doing.

You can explore our AI chatbot development services and AI customer service agents to understand how business-focused conversational systems can be designed around their intended use.

Even when a business chatbot is excluded, telling customers that they are interacting with AI remains a sensible transparency practice, and it keeps you ahead of any AI disclosure law that does apply.

If your application has both customer-service and companion-style functionality, do not assume the exclusion automatically covers every feature. Confirm its classification with qualified legal counsel.

Which AI tools should Colorado businesses review before 2027?

The following table is a practical starting point, not a legal classification of every possible implementation.

AI use in a small business Colorado ADMT law (SB 26-189)? Chatbot Safety Act (HB 26-1263)? What to do before 1 Jan 2027
Website support chatbot Likely excluded Likely excluded Check its primary purpose and document its intended use
Booking or scheduling agent Likely excluded Likely excluded Confirm it performs routine scheduling or transactional assistance
AI-written marketing and ad targeting Likely excluded Likely excluded Document marketing use and review data practices
AI résumé screening or candidate ranking Likely covered if it materially influences hiring Likely excluded Request vendor documentation and establish human review
AI pre-qualification for a loan or insurance quote Check with counsel; likely covered if it materially influences eligibility or access Likely excluded Review decision logic, notices and reconsideration procedures
AI tenant screening Likely covered if it materially influences housing decisions Likely excluded Establish notice, correction and meaningful human review
Public companion-style chat app Check with counsel based on its functionality Likely covered Review disclosure timing, minor-user safeguards and self-harm protocols

The distinction between screening and administrative support matters.

For example, a system that extracts information from résumés and organizes it for a recruiter may fall within an exclusion if it does not materially influence the hiring outcome.

A system that scores applicants and ranks them for selection presents a different situation.

Likewise, an AI tool that collects preliminary information for a loan inquiry is not necessarily performing the same function as a tool that evaluates eligibility and influences access to lending.

When classification is uncertain, ask qualified counsel rather than assuming that a vendor's marketing description determines whether the law applies. This article is general information, not legal advice.

How should a small business prepare for the Colorado AI Act?

You do not need to stop using AI or rebuild every workflow before January.

You need to understand where AI operates, what decisions it influences and who is responsible for reviewing its output.

Here is a practical preparation checklist.

Colorado AI readiness checklist

  • [ ] Inventory every AI tool. Include chatbots, CRM automations, scoring software, recruiting platforms, customer-service agents and third-party AI features.
  • [ ] Identify covered decision areas. Mark tools connected to employment, lending, insurance, housing, education, healthcare and essential government services.
  • [ ] Map the decision process. Determine whether AI only organizes information or generates scores, recommendations, rankings or classifications that influence outcomes.
  • [ ] Request vendor documentation. Ask developers for intended uses, data categories, limitations, known risks and instructions for monitoring and human review.
  • [ ] Prepare public notices. Write clear notices explaining when covered ADMT is used to materially influence consequential decisions.
  • [ ] Create an adverse-outcome template. Prepare a plain-language communication explaining the decision, the AI system's role, additional information requests and consumer rights.
  • [ ] Assign human review ownership. Name a responsible person who can investigate correction requests and reconsider decisions.
  • [ ] Set up record retention. Establish a process to retain relevant records for at least three years.
  • [ ] Review chatbot disclosures. Confirm whether publicly accessible conversational AI services fall within the Chatbot Safety Act and prepare disclosures where required.
  • [ ] Revisit the framework. Review your policies and vendor arrangements when the Attorney General publishes final rules.

One additional consideration: Colorado is not the only jurisdiction introducing AI-related obligations. Other states have their own AI and privacy rules. Businesses operating in the UK or Canada should also check the requirements applicable in their own jurisdictions.

Build AI governance into your existing operations

The most practical approach is to treat AI governance as part of normal software and business operations.

Every AI tool should have a defined purpose, an accountable owner, documented limitations and a clear escalation process.

For consequential decisions, human review should be designed into the workflow rather than added after something goes wrong.

At NextCTL, we approach AI implementation with this operational perspective. An AI agent should not simply produce an output and move on. Businesses need to know what it can do, where its authority ends and when a person must take over.

This is particularly important when connecting AI agents to CRM systems, customer records, hiring workflows and other business-critical applications.

Preparing your business for 2027? AI governance should help you deploy AI with clearer responsibilities, better oversight and more confidence. NextCTL LLC helps small and mid-sized businesses design, integrate and govern AI agents, automation systems and customer-facing AI applications. Book a free consultation with NextCTL to review your AI systems before 2027.

FAQ

Frequently Asked Questions

FAQ

When does the Colorado AI Act take effect?

The revised ADMT law under SB 26-189 and the Chatbot Safety Act under HB 26-1263 both take effect on 1 January 2027. The ADMT law applies to consequential decisions made on or after that date. Businesses should review their existing systems before implementation rather than waiting until the first covered decision occurs.

Does the Colorado AI Act apply to my customer service chatbot?

A typical customer-service chatbot is likely excluded from the Chatbot Safety Act because the legislation excludes services primarily designed for customer support and customer service. Routine customer-service triage is also excluded from the ADMT framework. However, functionality and primary purpose matter, so businesses should confirm their specific implementation with counsel.

What notice does a Colorado business need for AI hiring tools?

If an AI hiring tool materially influences a consequential employment decision, the deployer must provide clear and conspicuous notice before using it. Following an adverse outcome, the business must provide a plain-language description of the decision and the tool's role within 30 days, along with information about applicable rights. Consumers can request correction of inaccurate data and meaningful human review.

Does the Colorado law apply if my business is outside Colorado?

SB 26-189 defines a deployer as a person doing business in Colorado. Therefore, a business headquartered elsewhere may still need to examine its activities involving Colorado. The precise application depends on the business, its activities and the decisions involved. Businesses operating across state or national borders should check with qualified legal counsel.

Continue Reading

Independent check

Decide with evidence, not a pitch.

Open a neutral research prompt in the AI you already use. It asks where NextCTL fits, where a freelancer or an in house hire may be enough, and what to verify before you choose.

AI answers can be incomplete. Check nextctl.com/methodology before deciding.

See where your operations
are leaking revenue.

Tell us what you want to fix. We show where missed calls and slow follow up cost you most, with a written plan for what to fix first.

Or run a free tool first:Website AuditAI Readiness Scan