Colorado has enacted three laws regulating AI in healthcare this year: SB 26-189, which repealed and replaced much of Colorado’s landmark AI Act, as Sheppard has previously written on; HB 26-1139, regulating the use of AI in utilization review; and HB 26-1195 (the “Act”), concerning AI in psychotherapy services. HB 26-1195 went into effect on August 12 and establishes important new restrictions for Colorado psychotherapy providers seeking to use AI in their practices.
Permitted Uses of AI Under the Act
Under the Act, artificial intelligence systems may be used to provide administrative support or supplementary support so long as the licensed, registered, or certified individual maintains responsibility for reviewing all AI output.[1] “Administrative support” includes appointment scheduling and reminders, billing and insurance claims processing, and drafting general communications related to therapy logistics.[2] “Supplementary support” includes preparation and maintenance of client records (including therapy notes), analyzing data to track client progress or identify trends, organizing referrals, and collecting mental health or wellness information (including intake questionnaires).[3]
Prohibited Uses of AI Under the Act
The Act prohibits the use of AI to record or transcribe therapeutic sessions unless the provider has informed the client in advance that AI will be used, disclosed the specific purpose for the use of AI, and obtained written consent.[4] Adequate consent requires an unambiguous written agreement. The acceptance of general terms of use containing descriptions of AI along with other unrelated information, or interactions with digital content that involve hovering over, muting, pausing or closing digital content are not sufficient.[5]
Additionally, the Act prohibits AI from interacting with clients in any kind of therapeutic communication without synchronous, real-time interaction among the provider, the AI, and the client, and from generating therapeutic recommendations or treatment plans without the provider’s review and approval.[6]
Scope and Exceptions
The Act explicitly excludes guided meditation, mood monitoring, and other wellness tools that do not diagnose or treat mental health disorders so long as such tools clearly and conspicuously disclose that they are not a substitute for clinical care.[7]
The Act contains several exceptions to the prohibition. First, it allows licensed providers to use AI in accredited educational and training programs when such use is solely for educational, administrative, simulation, or training purposes and is not deployed for use with patients or the public.[8] Second, it permits providers to be involved in the development or testing of AI solely for research purposes under institutional review board oversight.[9] Third, the Act does not apply to providers who use wellness tools or FDA-authorized behavioral health technologies that provide the same clear and conspicuous disclosure.[10]
Advertising and Marketing Restrictions
Most of the Act focuses on its core AI-use framework. However, the Act also adds a companion provision to the Colorado Consumer Protection Act. This provision prohibits any person, not just regulated providers, from suggesting in AI advertising, interfaces, or outputs that the AI’s output is provided by, endorsed by, or equivalent to a licensed psychotherapy professional’s services or representing the AI system as providing psychotherapy services.[11]
Compliance Takeaways for a Multistate Landscape
Colorado was a leader in enacting its 2024 Artificial Intelligence Act, but other states have continued to regulate this space by adopting similar psychotherapy restrictions. HB 26-1195’s core provisions resemble those in Illinois’ legislation enacted last year, which Sheppard previously analyzed. Several states including Tennessee, Nevada, and Vermont have adopted advertising and marketing restrictions, some of which resemble those in the Act.
Despite these similarities, the patchwork of state AI laws presents an operational and compliance concern for any organization looking to deploy AI tools in the mental health space across multiple states. Psychotherapy providers operating in Colorado should consider:
- Confirming that a written, revocable client-consent workflow is in place for any AI-based session recording or transcription, triggered at initial use.
- Updating intake materials and initial-contact procedures to include the required written notice of the statute’s AI-use prohibitions.
- Auditing current and prospective AI tools against the statute’s three-way distinction between administrative support, supplementary support, and therapeutic communication. Tools that touch treatment planning, diagnosis, or direct client-facing therapeutic interaction may warrant the closest scrutiny.
- Reviewing marketing materials for language that could run afoul of the advertising restrictions, particularly claims implying that AI outputs are equivalent to licensed services or that an AI system is providing psychotherapy.
The mental health space has been an active ground for AI regulation. While state frameworks vary widely, many, like Colorado’s HB 26-1195, touch on common elements: (1) differing roles for AI in administrative and clinical functions; (2) the need for human review and approval; and (3) the need for regulation of marketing claims about the equivalence of AI output to human providers. These frameworks are likely to inform how lawmakers approach AI oversight in other areas of healthcare delivery. Flexible compliance programs that can accommodate a still-developing field are necessary to navigate the evolving regulatory landscape.
FOOTNOTES
[1] HB 26-1195 § 2(2).
[2] HB 26-1195 § 2(1)(a).
[3] HB 26-1195 § 2(1)(d).
[4] HB 26-1195 § 2(4)(a).
[5] HB 26-1195 § 2(1)(c).
[6] HB 26-1195 § 2(5).
[7] HB 26-1195 § 2(9)(a).
[8] HB 26-1195 § 2(7).
[9] HB 26-1195 § 2(8).
[10] HB 26-1195 § 2(9).
[11] HB 26-1195 § 3(1).