Back to Blog
AI

Colorado Repealed Its AI Law and Wrote a Different One. Your Clients Will Ask You About It.

15 September, 2026
4 min read
Colorado Repealed Its AI Law and Wrote a Different One. Your Clients Will Ask You About It.

Two states, two different answers, eighteen months apart.

What Colorado Now Requires

The Texas Responsible Artificial Intelligence Governance Act took effect on 1 January 2026. It restricts specific uses of AI the state treats as high risk, is enforced by the Texas Attorney General with civil penalties between $10,000 and $200,000 per violation, and carries no private right of action. It is intent-based: the question it asks is whether a system was deployed intentionally to discriminate, manipulate or harm.

Colorado went the other way and then changed its mind. A federal judge stayed enforcement of the original AI Act on 27 April 2026 after a legal challenge. On 14 May 2026 Governor Polis signed SB 26-189, repealing and replacing it with a narrower framework that takes effect on 1 January 2027.

The replacement is worth knowing in outline, because its shape is the one other states are likely to copy.

It covers automated decision-making technology used in consequential decisions, defined across education, employment, housing, financial services, insurance, health care and government services. Developers must give deployers documentation on intended uses, training data categories, limitations and instructions for meaningful human review. Deployers must give consumers clear and conspicuous notice, explain an adverse decision within 30 days, and provide meaningful human review on request. The Colorado Attorney General has exclusive enforcement authority and there is no private right of action.

Note what that is. It is not an accuracy standard and it is not a ban. It is a disclosure and explanation regime, and disclosure regimes are the ones small businesses fail by accident.

The Client Who Does Not Know They Are In Scope

Here is the conversation that is coming.

A client with forty employees uses a hiring platform with automated screening. Another uses a lending product that scores applicants. A third has a property manager running tenant applications through something. None of them describe what they do as deploying automated decision-making technology. All of them are, and none of them have read a word of this.

They will not call a law firm. They will ask their accountant, usually sideways, usually in the middle of a conversation about something else, and usually in the form of is this going to be a problem for us.

What to Actually Say

The honest answer has three parts and the first one matters most.

You are not their lawyer and this is not legal advice. Say it plainly and early, because the gap between a helpful observation and an advisory service they relied on is exactly where professional liability lives.

What you can do is help them find out whether the question applies. Which states do they operate in or hire in. Do they use any tool that makes or materially informs a decision about a person in employment, credit, housing, insurance or health care. Do they know whether a human reviews those decisions in a way that could be described as meaningful, and could they show it.

Then hand it on. A client who arrives at counsel with those answers gets a cheaper and better result than one who arrives with a question.

Why This Sits Close to the Books

The reason accountants get asked is not an accident of proximity.

The tools that make consequential decisions are usually the ones plumbed into payroll, billing and the general ledger, and the accountant is often the only outside professional who knows what software the business actually runs. That knowledge is the scarce part.

MetaWurks is built around that kind of retrieval. It ingests a client's contracts, invoices, statements and correspondence and lets an accountant query the whole set in plain English, so establishing which vendor agreements are in place and what they say is a question rather than a hunt through a shared drive. Documents ingested into the platform are not used to train models and are not exposed to other users, role based access controls decide who can open which client's records, and audit logs record who opened what and when.

It will not tell anyone whether a statute applies. It shortens the part where nobody can find the contract.

Two states have moved, one of them twice, and the federal position is unsettled. The value is not in knowing the statutes. It is in being the professional who noticed the client might be in one.

Join the Conversation

Do you know which of your clients uses software that makes or scores a decision about a person, and would they know to tell you?

Subscribe now to Our Newsletter and get the Coupon code.

All your information is completely confidential