Source: The National Law Review

Colorado’s HB 26-1195 establishes restrictions on AI in psychotherapy, according to legal analysis by Carolyn V. Metnick and Madeline Cunnion. The authors explain that administrative and supplementary AI support is permitted when the licensed, registered or certified professional remains responsible for reviewing all output. Examples include scheduling, billing, maintaining therapy records, tracking client progress, organizing referrals and collecting intake information. AI recording or transcription requires advance notification, disclosure of the specific purpose and the client’s unambiguous written consent. Acceptance of general terms containing unrelated information, or actions such as hovering over or closing digital content, does not provide sufficient consent. Therapeutic communication involving AI requires synchronous, real-time interaction among the provider, AI and client. AI-generated therapeutic recommendations and treatment plans require provider review and approval.

The analysis describes an exclusion for meditation, mood monitoring and other wellness tools that neither diagnose nor treat mental disorders and clearly disclose that they do not replace clinical care. An exception covers licensed providers’ AI use in accredited educational and training programs solely for educational, administrative, simulation or training purposes, without deployment to patients or the public. Another exception permits provider participation in AI development or testing solely for research under institutional review board oversight. The authors also identify an exclusion for providers using wellness tools or FDA-authorized behavioral health technologies carrying the required clear disclosure. A companion consumer-protection provision restricts any person from presenting AI as providing psychotherapy or suggesting that AI output comes from, is endorsed by or is equivalent to a licensed psychotherapy professional’s services. The authors recommend revocable written-consent procedures, required written notices, audits of AI functions and reviews of marketing claims. Metnick and Cunnion argue that differing state requirements complicate multistate compliance and predict that these frameworks may influence AI oversight elsewhere in healthcare.