Published 1 hour ago

Colorado AI Healthcare Law Takes Effect: New Guardrails for Mental Health Chatbots

Colorado just drew a clear line: a chatbot is not your therapist. Starting August 12, HB26-1195 makes that official.

The law targets a real and growing problem — AI tools marketed to patients as legitimate mental healthcare. Some have implied HIPAA protections they don’t actually provide. Others have, in documented cases, given advice that ended lives. Colorado’s response isn’t a blanket ban on AI in healthcare. It’s a boundary.

3 mins read

6 sections

Key Highlights

Colorado law now bans marketing chatbots as equivalent to licensed therapists
HB26-1195 allows administrative AI in behavioral health but under clear disclosure rules
The law offers a compliance checklist for mental health AI tools used with Colorado patients

What the Law Actually Does

HB26-1195 is more surgical than sweeping. Here’s the short version:

Therapy must be human-delivered. Psychotherapy and clinical treatment require a licensed professional — a psychologist, social worker, addiction counselor, or equivalent.
Administrative AI is still allowed. Transcription, scheduling, session notes — AI can assist with these, under licensed oversight.
Disclosure is required. Providers must tell patients when AI tools are being used in a supplementary capacity.
No fake equivalence. Chatbots cannot be marketed as equivalent to a licensed therapist or counselor.
No false HIPAA claims. AI tools cannot imply patient conversations carry standard confidentiality protections when they don’t.

The distinction matters. The law isn’t anti-AI. It’s anti-deception.

Why This Became Necessary

The context here is hard to ignore. Major AI companies — including OpenAI, Google, and Character.AI — are facing lawsuits from families after chatbots provided harmful responses to people in crisis. Parents testified before Congress about teens who died by suicide after interacting with AI tools that discouraged them from seeking real help.

Stanford researchers concluded in 2025 that large language models “should not replace therapists” and found that LLMs express stigma toward people with mental health conditions and respond poorly to critical situations.

That’s not a fringe finding. That’s a structural problem with using general-purpose language models in high-stakes clinical contexts.

The Compliance Angle for AI Tool Builders

If you’re building or selling AI tools in the behavioral health space, Colorado’s law sets a concrete compliance checklist — even if you’re not based there.

Things that now require attention:

Marketing copy. Does your product description imply therapeutic equivalence? That’s now legally problematic in Colorado.
Confidentiality language. Are you suggesting HIPAA-level protections your tool doesn’t actually provide? That needs to change.
Disclosure flows. If your tool is used in a clinical setting, the provider needs a clear way to inform patients.

The law applies to tools used with Colorado patients, not just companies headquartered there.

The Broader Signal

Colorado isn’t operating in a vacuum. This is part of a visible pattern — states moving ahead of federal policy to set specific rules for AI in sensitive domains. Mental health is an obvious starting point: high vulnerability, high stakes, and a market that moved fast without guardrails.

The administrative AI carve-out is notable. Lawmakers aren’t trying to freeze AI out of healthcare. They’re trying to keep it in its lane — handling paperwork, not people in crisis.

For anyone evaluating AI tools in healthcare contexts, HB26-1195 is a useful framework regardless of where you operate. The questions it forces — Is this tool being positioned as clinical care? Does it make confidentiality claims it can’t back up? Is there licensed human oversight? — are good questions everywhere.

The law takes effect August 12. If your stack touches behavioral health in Colorado, that’s your deadline.