House Bill 26-1139 regulates the use of artificial intelligence in Colorado health care, with a focus on utilization review and psychotherapy. For utilization review, the bill applies to health insurers, pharmacy benefit managers, private utilization review organizations, behavioral health administrative services organizations, and managed care entities that use AI to help decide whether health-care services are covered. It requires AI systems to rely on an individual patient’s medical and clinical information rather than group data alone, to avoid discriminatory use, to maintain documentation and audit logs, and to be periodically reviewed for accuracy and reliability. The bill also requires written disclosures to state agencies describing how AI is used and what human oversight exists.
The bill also addresses AI used in mental health care. It defines a “mental health companion chatbot” and treats certain AI representations or conduct as unauthorized practice of psychotherapy if the system presents itself as a human provider, uses protected professional titles, delivers psychotherapy without human oversight, or fails to give clear notice that it is not a human therapist. Such chatbots must disclose that they are AI when asked, provide crisis-response protocols for suicidal ideation or self-harm, and may not sell or share identifiable mental health data without consent. The bill further prohibits health insurers, Medicaid-related payers, and the Children’s Basic Health Plan from paying for psychotherapy services that are directly delivered by AI systems.
In state law, the bill adds new provisions to the insurance code and the Medicaid/CHP+ payment statutes, creating explicit limits on automated coverage decisions and on reimbursement for AI-delivered psychotherapy. It preserves the use of AI as an assistive tool, including for expedited approvals and administrative functions, but bars adverse medical-necessity denials from being issued solely on AI output without review and approval by a qualified licensed clinician, physician, or other regulated professional. It also requires regulated professionals to disclose to clients when they use AI systems or AI-enabled diagnostic or therapeutic devices in practice.
The overall sentiment reflected in the bill text is strongly supportive of human-centered health care and cautious about AI. The legislative declaration emphasizes compassion, individualized clinical judgment, equity, and patient safety, while acknowledging that AI can be useful for efficiency and support functions. The bill’s structure suggests a policy preference for allowing AI as a tool, but not as a substitute for licensed professional judgment in coverage decisions or psychotherapy.
The main points of contention are likely to center on the scope of AI restrictions, the burden of compliance and documentation, and whether the bill goes too far in limiting innovation in health care delivery. Potentially affected parties include insurers, pharmacy benefit managers, utilization review vendors, behavioral health organizations, Medicaid and CHP+ administrators, and providers using AI-enabled tools. The strongest policy concern on the bill’s supportive side is preventing biased or opaque automated denials and protecting patients from AI posing as mental health professionals or providing therapy without adequate safeguards.
The bill amends Colorado insurance and public health-care payment laws to require human oversight in AI-assisted utilization review and to prohibit payment for psychotherapy services directly delivered by AI systems. It creates new disclosure, documentation, and review obligations for carriers and related entities using AI in coverage determinations, and it bars Medicaid/CHP+ payers and private carriers from reimbursing AI-conducted psychotherapy. It also adds professional-practice disclosure requirements for regulated clinicians who use AI tools in treatment or diagnosis.
The bill’s tone and stated findings are generally favorable toward regulating AI rather than banning it outright. The discussion embedded in the bill text frames AI as useful for administrative efficiency but insufficient for making or materially influencing adverse health-care decisions without human judgment. The overall sentiment is protective of patients and clinicians, with a clear preference for transparency, accountability, and human review.
The likely areas of contention are whether AI should be allowed to make or heavily influence utilization-review decisions, how much human review is enough, and whether the bill’s restrictions could slow prior authorization or increase administrative costs. Another likely point of debate is the treatment of mental health companion chatbots: supporters would view the notice, crisis-response, and data-use limits as necessary consumer protections, while critics may argue they are too restrictive for emerging digital mental-health tools. Insurers, PBMs, utilization review organizations, and AI vendors are the parties most likely to resist the compliance and reimbursement limits, while patient advocates, clinicians, and mental health regulators are likely to support them.