Picture an apartment application in Lakewood. A screening company runs your rental history, court records and credit through its software, and you come back as a bottom-quartile risk. The landlord says no. You pay the application fee again at the next place, and the place after that, and nobody ever tells you a score was involved.
Colorado decided that shouldn't happen anymore. This spring the legislature passed two laws about artificial intelligence, and the governor signed both on May 14, 2026. They take effect on January 1, 2027.
Written comments on the rules close at 11:59 p.m. Mountain. The public hearing is the same day at 10:00 a.m. at the Colorado Department of Law in Denver, with a Zoom option.
Leave a comment at coag.gov/ai ↗ How to do it in 10 minutesThe two laws
The Automated Decision-Making Technology Act (SB 26-189) covers what lawyers call consequential decisions: housing, jobs, credit, health care, education. When an automated system plays a real part in a decision that goes against you, you are owed a notice. That notice has to say an automated system was used, give the main reasons, tell you what data it relied on, explain how to correct that data, and tell you how to get a human to review it.
The Conversational Artificial Intelligence Services Act (HB 26-1263) covers the chatbots people talk to. Its rules are being written in the same process. Our comment focused on the first law, because that is where a person's housing, paycheck or care is on the line.
The Attorney General enforces both. There is no private right to sue under the decision law, which means the rules the Attorney General writes, and the records they require companies to keep, are most of what stands behind your rights.
The three fights over the rules
The law sets the outline. The Attorney General's office is now filling in the details, in a rule set called 4 CCR 904-6. Most of the comments on file come from the companies that will have to follow it. Here is what they are pushing on.
1. When does the law even apply?
The law kicks in when an automated system "materially influences" a decision. The draft offers two ways to read that. The first counts any real influence that isn't trivial. The second only counts it when the AI outweighed everything else. Under the second reading, the landlord in our example points to a gap in your rental history, says that mattered more, and you get no notice at all. Same software. Same denial. No rights.
2. Do they have to name who supplied the data?
The draft says a company must name each actual source of the data used against you: the specific data broker, court, database or employer. Some commenters want to replace names with categories like "a consumer reporting agency" or "public records." You can't send a correction request to a category. Federal regulators found that 1 in 5 people had an error on at least one credit report, and 4 out of 5 who disputed an error got their report changed. People win when they know who to write to.
3. Is the human review real?
The draft says the person reviewing your case can't be a subordinate of whoever made the original call, can't be punished for overruling it, and can't hand the review back to another AI. Some commenters want all three removed. A review by someone who reports to the original decision-maker isn't really a second look.
There is also a timing fight. Some companies want a grace period that would push real protection past January 1, 2027, and some want older federal credit notices to count as good enough. The legislature already set the start date.
What we asked for
On October 2, The Q Collective filed a 9-page comment asking the Attorney General to keep the first reading of "materially influence," require named data sources, keep human review independent and fully human, keep two-way notice delivery and a 10-business-day answer time, and hold the January 1 start date. We have no client and no money in the outcome.
How to weigh in, in about 10 minutes
- Open the comment form at coag.gov/ai. It asks for your name, email, phone, what kind of commenter you are, and any organization you're with. "Individual" or "Colorado resident" is fine.
- Pick the topic. Choose the Automated Decision-Making Technology Act.
- Write three to five sentences. Who you are, why it matters to you, and which rule you want kept. A real story beats a long argument every time. You can type it into a document and upload it, since the form takes file uploads.
- Know that it's public. The Attorney General posts comments online, so leave out anything you wouldn't want read aloud.
- Submit by 11:59 p.m. Mountain on Monday, October 26.
Here's a starting point. Change it into your own words:
Want to be there in person?
The hearing starts at 10:00 a.m. Monday, October 26, at the Colorado Department of Law, Ralph L. Carr Judicial Building, 1300 Broadway, Denver. You can also join by Zoom. The registration link is posted with the hearing notice. If you can't find it, call the Attorney General's office at (720) 508-6000 and ask for the ADMT rulemaking hearing.
What happens next
After October 26 the Attorney General's office writes the final rules. On January 1, 2027, the law takes effect, and the first notices should start reaching people who get turned down. We'll keep track of all of it on our timeline and in the Sunday Update.
Sources: SB 26-189 and HB 26-1263; Colorado Department of Law, proposed rules 4 CCR 904-6 (August 11, 2026) and comment page; Federal Trade Commission, Report to Congress under Section 319 of FACTA (credit report accuracy). Our comment cites each in full. Spot an error? Tell us and we'll correct it in public.