Colorado's AI-Therapy Real-Time Monitoring Law: What AI Builders Need to Know

Colorado has enacted a law that requires therapists using AI in psychotherapy to be actively logged into the AI system and monitor client-AI interactions in real time. HB1195, signed on June 3, 2026 and codified as Section 12-245-224.5, includes a provision (part 5a) that mandates synchronous, real-time, interactive co-participation by the therapist with the client and AI. While intended as a safety measure, the requirement creates significant practical burdens for therapists and raises questions about reimbursement, liability, and the future of AI in mental health care.

What happened

Colorado's HB1195 requires that when a therapist uses AI as a client-facing therapeutic tool, the therapist must co-participate in the AI tool usage with the client on a synchronous, real-time, interactive basis. The therapist must be logged into the AI, paying attention, and ready to intervene instantly.

The law also requires written, informed consent from clients if AI will record or transcribe sessions. Generation-based documentation tools are explicitly exempt as supplementary support.

Why AI builders should care

For teams building AI-powered mental health tools, this law introduces a regulatory constraint that could reshape product design and adoption. The real-time monitoring requirement may limit how AI tools integrate into therapy workflows, potentially affecting reimbursement models and therapist time valuation.

The law appears to treat all AI as equally risky, making no distinction between general-purpose AI (GPAI) and purpose-built AI (PBAI) for mental health. This could slow adoption of specialized, vetted AI tools and push clients toward unsupervised AI use to avoid monitoring costs.

Practical implications

Therapists may avoid AI tools altogether to sidestep the time burden and potential malpractice risk. If a client later claims the AI gave harmful advice and the therapist failed to intervene, the law creates a clear legal duty. This liability exposure could discourage adoption even among therapists who see value in AI.

Clients may choose to use AI on their own without telling their therapist, creating a gap between professional guidance and unsupervised AI use. This could undermine care continuity and safety.

For AI builders, the law signals a need to design tools that support compliant workflows without overburdening clinicians. Features like automated alerts for concerning AI outputs, session summaries for therapist review, and clear consent workflows could help.

Caveats

The analysis relies on public commentary and bill text. The exact regulatory language and scope may be clarified by official summaries and enforcement guidance. The law includes a potential loophole: if the AI is not purporting to diagnose or treat mental health disorders, the real-time monitoring requirement may not apply. AI makers may use this distinction to position tools as supporting mental well-being rather than providing therapy.

This law is one of several state-level AI mental health regulations, and the broader federal-state legal dynamics remain unresolved. Builders should monitor how these laws interact and whether federal AI legislation eventually preempts state rules.