Your Therapist Can't Be a Bot: Inside Colorado's New AI Health Care Bill

The Bottom Line

Insurance companies are increasingly using artificial intelligence to process and deny medical claims, and AI chatbots are starting to act a lot like real therapists. This bill pumps the brakes by forcing a human doctor to review any AI-recommended health coverage denial, strictly regulating mental health chatbots, and banning insurers from paying for AI-led therapy. It's all about making sure a computer algorithm isn't making the final call on your physical or mental health.

What This Bill Actually Does

At its core, this bill tackles the growing friction between automated technology and human-centered health care. First, it addresses utilization review — the process health insurance companies use to decide whether they will pay for a specific medical treatment or service. Under Section 2 of the bill, insurance carriers and pharmacy benefit managers can still use an artificial intelligence (AI) system to speed up paperwork and approve your care. However, if the AI flags a claim to be delayed or denied based on "medical necessity," the computer cannot make the final call. A licensed clinician or physician who actually understands the specific medical issue must independently review and approve the denial. Furthermore, the AI can't just rely on broad statistical "group data" to deny a claim; it has to factor in your individual clinical history.

Second, the legislation creates strict guardrails for a new category of technology: the mental health companion chatbot. These aren't just generic guided meditation apps. The bill defines them as systems using generative AI to have dynamic, personalized, back-and-forth conversations that mimic real therapy. The state isn't banning these bots, but it is requiring radical transparency. If an AI acts like a therapist, it must constantly remind users (via unskippable pop-ups every 30 minutes) that it is a machine and not a licensed professional. If a user expresses suicidal thoughts or self-harm, the AI is legally required to trigger a protocol referring the user to a crisis text line or suicide hotline. Crucially, the bill makes it illegal for these tech companies to sell or share a user's identifiable mental health data.

Finally, the bill draws a hard line on health care billing. It strictly prohibits therapists from billing a patient, a private insurance company, or public programs like Medicaid for therapy sessions conducted by an AI. It also requires human therapists to be completely upfront with their clients. If your therapist uses ambient AI to take session notes, or uses an AI-powered diagnostic tool, they must disclose the name of the software, how your privacy is protected, and give you the right to opt out of having your confidential communications fed into an algorithm.

What It Means for You

If you've ever had a doctor order a vital test like an MRI, only to have your insurance company automatically deny it, this bill changes the game. Starting January 1, 2027, you won't be trapped in an endless loop fighting a computer algorithm that rejected your claim just to save a few bucks. Any denial based on medical necessity must have a real, human expert's review attached to it. While this doesn't guarantee your insurance will approve every treatment, it does guarantee that your unique medical history was actually considered, rather than your claim being tossed out because you fell into an unfavorable statistical bucket.

If you or your kids use AI companion apps to cope with stress or anxiety, your privacy is getting a massive upgrade. The bill makes it illegal for these tech companies to sell your identifiable mental health data to advertisers or data brokers. They also can't force you to consent to data sharing just to access the app. Plus, if things get serious and a user mentions self-harm, you can trust that the app is legally obligated to immediately pause the chat and connect the user with a real human crisis hotline.

For those who see a human therapist, you might start noticing some new paperwork in your intake packets. Because so many behavioral health providers are adopting AI tools to help with note-taking and diagnostics, your provider will now be legally required to disclose exactly how they use AI during your sessions. It is highly recommended that you read these new disclosures carefully. You'll have the power to see exactly what software is being used, whether it is FDA-approved, and you will retain the right to explicitly consent (or refuse) to have your most vulnerable, confidential conversations analyzed by an artificial intelligence system.

What It Means for Your Business

For health insurance carriers, pharmacy benefit managers (PBMs), and private utilization review organizations, the operational shifts required by the January 1, 2027 effective date are substantial. You must build "human-in-the-loop" workflows for any coverage denial based on medical necessity. You are also required to submit written disclosures to state regulators — like the Division of Insurance or the Department of Health Care Policy and Financing — detailing how your AI systems are governed. Importantly, you must maintain robust audit logs to prove your algorithms are not resulting in unfair discrimination, and you must ensure your AI systems are evaluating individual clinical records, not just denying claims based on aggregate group data.

If you are a tech developer or a startup building health apps or mental health companion chatbots, your product roadmap just inherited strict compliance mandates. Your user interface must feature specific, recurring disclosures — including a pop-up window every 30 minutes explicitly stating the system is not a human and not a licensed provider. You must build hardcoded pathways to suicide hotlines for any user input mentioning self-harm. Most importantly, your business model cannot rely on monetizing user mental health data; selling or sharing this data is prohibited. (Note: There is a safe harbor exemption if you are developing and testing these systems solely for research under the oversight of an Institutional Review Board registered with the federal government).

For licensed therapists, social workers, and behavioral health clinics, the billing rules are absolute: you cannot bill public or private payers for AI-conducted therapy, AI-led supervision of your clinical candidates, or AI professional consultations. Additionally, you need to update your client intake forms and consent waivers immediately. If you use AI-assisted therapeutic devices, diagnostic tools, or even ambient AI scribes to write your clinical notes, you must provide a written disclosure to your clients. This must include the specific name of the AI, a website where the tool is described, proof of how you are protecting HIPAA data (such as having a Business Associate Agreement in place with the software vendor), and a clear mechanism for the client to consent to or decline the use of the tool.

Follow the Money

Despite the heavy regulatory framework and the creation of entirely new consumer protections, this bill doesn't come with a massive price tag for taxpayers. According to the state's fiscal note, the legislation requires no new state appropriations.

The regulatory burden falls primarily on the Division of Insurance within the Department of Regulatory Agencies (DORA), which will need to issue new rules and review the incoming AI disclosures from insurance carriers. However, the state expects to handle this workload within its existing operational budget. There may also be a very slight uptick in civil case filings if patients sue over discriminatory AI practices or unauthorized psychotherapy, generating a tiny bit of court revenue, but the overarching financial impact to the state's general fund is functionally zero.

Where This Bill Stands

HB26-1139 is currently Signed Into Law. The latest official action came on 06/02/2026: Governor Signed.

That means the legislative process is complete and the bill is now law. The remaining questions are about implementation timing and how agencies, businesses, or local governments respond.