Colorado Psychotherapy AI Restrictions Act

Overview

Colorado House Bill 1195 (HB 26-1195), officially titled "PSYCHOTHERAPY ARTIFICIAL INTELLIGENCE RESTRICTIONS," is a landmark piece of legislation in the United States, specifically within the state of Colorado, designed to regulate the integration and use of artificial intelligence (AI) systems within the provision of psychotherapy services. Enacted on June 3, 2026, and slated to take effect on August 12, 2026, this Act establishes critical guardrails to ensure that mental healthcare remains human-centered and that patients are protected from the potential risks associated with unregulated AI deployment in sensitive therapeutic contexts. The core intent of the legislation is to prohibit AI systems from directly engaging in therapeutic communication, generating unsupervised treatment recommendations, or detecting emotions or mental states without direct human professional oversight. This proactive stance by Colorado aims to balance the potential benefits of AI for administrative and supplementary support with the imperative to maintain the ethical and professional standards of mental health care.

The Act underscores a fundamental principle that the expertise, nuance, and interpersonal connection inherent in psychotherapy cannot be replicated by AI-generated algorithms. It explicitly states that clinical treatment and psychotherapy must be administered by a licensed human professional, thereby safeguarding the quality and integrity of care. While the bill acknowledges the utility of AI for administrative functions, such as recording or transcribing sessions, it mandates strict conditions for such use, including the requirement for written informed consent from clients and the retention of full professional responsibility by the regulated mental health professional. Furthermore, HB 1195 introduces significant enforcement mechanisms, classifying violations as unfair trade practices under the Colorado Consumer Protection Act and empowering relevant licensing boards within the Department of Regulatory Agencies (DORA) to impose disciplinary actions and substantial penalties. This comprehensive approach positions Colorado at the forefront of state-level efforts to responsibly govern AI in healthcare, particularly in the sensitive domain of mental health.

Definitions

Central to the understanding and implementation of Colorado HB 1195 are several key definitions that delineate the scope and application of the Act. An "Artificial Intelligence (AI) system" is broadly understood as any system designed to perform tasks that typically require human intelligence, such as learning, problem-solving, and decision-making, particularly when applied in the context of psychotherapy services. The legislation does not seek to ban AI outright but rather to define its permissible and impermissible uses within a clinical setting. "Psychotherapy services" are defined in accordance with existing state law, encompassing treatment, diagnosis, testing, assessment, or counseling within a professional relationship aimed at assisting individuals or groups to alleviate behavioral and mental health disorders, understand motivation, resolve emotional conflicts, or modify behaviors. This definition is crucial as it clarifies the specific professional activities that fall under the purview of the new AI restrictions.

The Act also precisely defines who constitutes a "Regulated professional," referring to licensed, certified, or registered psychologists, professional counselors, social workers, clinical social workers, marriage and family therapists, addiction counselors, or candidates for these credentials, as well as unlicensed psychotherapists and other individuals lawfully permitted to provide psychotherapy services in Colorado. This broad inclusion ensures comprehensive coverage across the mental health spectrum. "Therapeutic communication" is identified as direct interaction with clients intended to facilitate psychological or emotional healing and growth, an area where AI is explicitly restricted from independent engagement. Similarly, "Treatment recommendations" refer to suggestions or plans for therapeutic interventions, which AI systems are prohibited from generating without review and approval by a regulated professional. The concept of "Informed consent" is vital, requiring clients to be fully informed about the use of AI, including for administrative or supplementary support, and to provide voluntary written agreement. These definitions collectively establish a clear framework for ethical and compliant AI integration in mental health care, emphasizing human oversight and patient autonomy.

Governance and Institutional Framework

The governance and institutional framework for enforcing Colorado HB 1195 primarily rests with the Department of Regulatory Agencies (DORA) and its associated licensing boards. DORA is the overarching state department responsible for regulating various professions and occupations in Colorado, including those within the mental health sector. Under this Act, the applicable licensing boards, such as the State Board of Psychologist Examiners, the State Board of Licensed Professional Counselors, or the State Board of Marriage and Family Therapist Examiners, are explicitly authorized to take disciplinary action against regulated professionals who violate the provisions concerning the use of AI systems in psychotherapy. This delegation of authority to existing professional regulatory bodies leverages their expertise in professional standards and ethical conduct, ensuring that enforcement is carried out by entities familiar with the nuances of mental health practice. The Act integrates AI regulation directly into established professional oversight mechanisms, rather than creating an entirely new regulatory body.

Beyond professional disciplinary actions, the Act also establishes a broader enforcement mechanism through the Colorado Consumer Protection Act (CCPA). This means that violations related to AI advertising and restrictions can be pursued as unfair trade practices, involving the Department of Law and potentially district attorneys. This dual approach provides a robust framework for both professional accountability and consumer protection. The involvement of the CCPA extends the reach of the Act beyond individual practitioners to any person, corporation, or entity that might unlawfully provide or advertise psychotherapy services through AI. The fiscal notes associated with HB 1195 indicate that DORA will experience a minimal increase in workload for education, outreach, and complaint response, and may require legal services from the Department of Law for rulemaking and implementation. This suggests an ongoing role for DORA in developing specific rules and guidelines to operationalize the Act's provisions, ensuring clarity and consistency in its application across regulated mental health professions.

Key Focus Areas

Colorado HB 1195 zeroes in on several critical aspects of AI integration in mental health services, primarily focusing on maintaining human oversight and patient safety. A paramount focus area is the outright prohibition of AI systems from directly engaging in therapeutic communication with clients. This means AI cannot independently conduct therapy sessions, provide counseling, or interact with patients in a manner that constitutes direct clinical intervention. The rationale behind this restriction is to preserve the essential human element of empathy, judgment, and connection that is fundamental to effective psychotherapy. This prohibition extends to preventing AI from generating therapeutic recommendations or treatment plans without thorough review and explicit approval by a regulated professional, ensuring that clinical decisions remain the responsibility of a qualified human expert.

Another significant focus is the restriction on AI systems detecting emotions or mental states. This provision addresses concerns about the accuracy, bias, and ethical implications of AI attempting to interpret complex human emotional and psychological states, which often require nuanced contextual understanding and professional training. While direct therapeutic engagement and independent diagnostic functions are restricted, the Act does permit the use of AI for administrative or supplementary support functions. This includes tasks such as scheduling, billing, or transcribing therapeutic sessions, provided that the regulated professional maintains full responsibility for all AI interactions, outputs, and data use. Crucially, if AI is used to record or transcribe therapeutic sessions, the Act mandates that written informed consent must be obtained from the client. This requirement highlights the Act's commitment to transparency and patient autonomy, ensuring individuals are fully aware and agree to how technology is utilized in their care. The legislation also carves out exceptions for the use of AI in accredited educational and training programs, as well as for research purposes under the oversight of a federally registered institutional review board, acknowledging the importance of innovation and professional development in controlled environments.

Implementation Framework

The implementation framework for Colorado HB 1195 is designed to integrate the new AI restrictions into existing regulatory structures for mental health professionals. Upon its effective date of August 12, 2026, the Act will become binding law, requiring immediate adherence from all licensed, certified, or registered psychotherapy providers in Colorado. The Department of Regulatory Agencies (DORA) and its specific licensing boards for mental health professions are tasked with the primary responsibility for overseeing compliance and enforcing the Act's provisions. This involves conducting education and outreach initiatives to inform mental health professionals about the new requirements and prohibitions. DORA's role will also extend to responding to complaints regarding AI use and potentially engaging in rulemaking processes to provide more detailed guidance on the application of the Act. The Behavioral Health Administration (BHA) is also expected to have a minimal increase in workload to update its rules to align with the new legislation, ensuring a cohesive regulatory environment.

Furthermore, the Act's integration with the Colorado Consumer Protection Act (CCPA) provides an additional layer of implementation and enforcement. The Department of Law and district attorneys will be involved in pursuing cases where individuals or entities engage in deceptive trade practices related to the advertising or provision of psychotherapy services through AI systems. This dual enforcement mechanism ensures that both professional conduct and broader consumer protection are addressed. The fiscal notes associated with the bill anticipate a minimal increase in workload for the Judicial Department due to potential civil case filings under the CCPA, although it is assumed that most professionals will comply with the law. The Act's carve-outs for educational, administrative, simulation, or training purposes, and for research programs under institutional review board (IRB) oversight, are crucial components of the implementation framework, allowing for responsible innovation and development of AI tools within controlled environments without impinging on patient safety in clinical settings. These exceptions are vital for fostering responsible AI advancement while maintaining strict patient protections in direct care.

Monitoring and Evaluation

The monitoring and evaluation of Colorado HB 1195 will primarily be conducted through the existing oversight mechanisms of the Department of Regulatory Agencies (DORA) and its associated licensing boards. These bodies are responsible for receiving and investigating complaints related to professional conduct, and the Act explicitly grants them the authority to take disciplinary action against regulated professionals who violate its provisions regarding AI use. This ongoing complaint-driven monitoring will serve as a key indicator of compliance and potential areas where further clarification or enforcement may be needed. DORA’s anticipated increase in workload for education, outreach, and complaint response suggests an active role in observing the practical application of the Act and identifying any emerging issues. The effectiveness of these monitoring efforts will depend on the vigilance of both the public and professionals in reporting non-compliant AI use, as well as the resources allocated to DORA for investigation and enforcement.

Beyond individual complaints, the broader impact of the Act on consumer protection will be monitored through the Colorado Consumer Protection Act (CCPA). The Department of Law and district attorneys will track instances of deceptive trade practices related to AI in psychotherapy, which could provide insights into market trends and the prevalence of unlawful advertising or service provision by AI systems. While the fiscal notes project a minimal increase in civil case filings, any significant rise could indicate a need for re-evaluation of the Act's deterrent effects or the clarity of its provisions. Over time, the data collected from disciplinary actions, civil penalties, and public complaints will be crucial for evaluating whether the Act is effectively achieving its goals of safeguarding patient care and ensuring responsible AI integration. This continuous feedback loop will be essential for identifying whether the current regulatory framework is sufficient or if amendments, additional guidance, or increased enforcement efforts are required to adapt to the evolving landscape of AI in mental health services.

Penalties, Liability, and Appeals

Colorado HB 1195 establishes clear penalties and mechanisms for liability and appeals to ensure compliance and accountability in the use of AI in psychotherapy services. For licensed, certified, or registered mental health professionals who violate the Act's provisions, the applicable licensing boards within the Department of Regulatory Agencies (DORA) are authorized to take disciplinary action. This can include administrative penalties of up to $5,000 for each violation, with the revenue from these fines credited to the respective licensing board cash funds. These administrative fines serve as a direct deterrent for individual practitioners and underscore the professional responsibility associated with AI use. The disciplinary actions can range from reprimands and license suspensions to revocation, depending on the severity and frequency of the violations, aligning with existing professional conduct regulations.

In addition to administrative penalties, the Act leverages the Colorado Consumer Protection Act (CCPA) to address broader instances of non-compliance. It makes it unlawful for any individual, corporation, or entity to provide, advertise, or otherwise offer psychotherapy services to the public in the state, including through the use of an AI system, unless those services are provided by a regulated professional. Violations of these provisions, particularly those involving deceptive advertising or misrepresentation of AI as legitimate psychotherapy, constitute an unfair trade practice under the CCPA. Individuals or entities found to be committing such deceptive trade practices may be subject to a civil penalty of up to $20,000 for each violation, with additional penalties for subsequent violations of a court order or injunction. These civil penalties are classified as damage awards and are not subject to the Taxpayer's Bill of Rights (TABOR). The Department of Law and district attorneys are empowered to pursue these civil actions. The appeals process for disciplinary actions by licensing boards typically follows established administrative procedures, allowing regulated professionals to challenge findings and penalties. Similarly, civil penalties under the CCPA would be subject to the standard judicial appeals process, ensuring due process for all parties involved.

Relationship to Other Instruments

Colorado HB 1195 operates within a broader legal and regulatory landscape, interacting with several existing instruments at both the state and federal levels. Primarily, it builds upon and amends the existing Mental Health Practice Act in Colorado, integrating specific restrictions and requirements for AI use directly into the framework governing mental health professionals. This ensures that AI regulation is not a standalone measure but an embedded component of professional standards and ethical practice. The Act also explicitly references and utilizes the Colorado Consumer Protection Act (CCPA), classifying certain violations related to AI advertising and the unlawful provision of psychotherapy services as unfair trade practices. This connection broadens the enforcement scope, allowing for civil penalties and legal action beyond professional disciplinary measures, thereby leveraging an established consumer protection framework to address emerging AI-related concerns.

Furthermore, the Act emphasizes compliance with existing state and federal privacy and security laws in its provisions regarding AI use for administrative or supplementary support. This implicitly refers to statutes such as the Health Insurance Portability and Accountability Act (HIPAA) at the federal level, which sets national standards to protect sensitive patient health information, and potentially state-level privacy laws. By requiring AI use to adhere to these established privacy frameworks, HB 1195 reinforces the importance of data protection in the context of AI in healthcare. The legislation also has a nuanced relationship with other AI-related bills in Colorado, such as SB 26-189 (the Colorado AI Act) and HB 26-1139 (focused on utilization review in mental health). While SB 26-189 provides broader AI governance, HB 1195 is specifically tailored to the unique challenges and sensitivities of psychotherapy, offering more granular and direct prohibitions. This indicates a legislative strategy where general AI governance is complemented by sector-specific regulations to address particular risks in critical domains like mental health.

National/Federal Alignment

Colorado HB 1195 represents a significant state-level initiative in the United States to regulate artificial intelligence, particularly within the sensitive domain of mental health services. At the national level, the United States currently lacks a comprehensive federal AI regulation framework akin to the European Union's AI Act. Instead, federal efforts have largely focused on guidelines, executive orders, and sector-specific approaches. For instance, the National Institute of Standards and Technology (NIST) has developed an AI Risk Management Framework, which provides voluntary guidance for managing risks associated with AI. While not a binding regulation, such federal frameworks offer principles that state-level legislation like HB 1195 can align with, particularly in areas of risk assessment, transparency, and accountability. The Colorado Act's emphasis on professional responsibility, informed consent, and the prohibition of unsupervised AI clinical interaction aligns with the broader ethical principles often discussed in federal AI policy dialogues, which prioritize human oversight and safety in high-risk applications.

In the absence of a direct federal mandate for AI in mental health, state laws like HB 1195 play a crucial role in establishing specific guardrails tailored to local needs and existing state regulatory structures. The Act's requirement for compliance with state and federal privacy and security laws, such as HIPAA, demonstrates its integration into established federal legal protections for health information. This ensures that while Colorado is innovating in AI regulation, it is doing so within the confines of existing federal patient data privacy standards. The state-specific nature of professional licensing and consumer protection laws (like the Colorado Consumer Protection Act) means that enforcement mechanisms are inherently localized. However, as other states consider similar legislation, Colorado's approach could serve as a model or influence national discussions on how to effectively regulate AI in healthcare without stifling innovation. This fragmented regulatory landscape highlights the ongoing challenge of achieving national alignment on AI governance, with states often leading the way in addressing specific sectoral concerns.

Implementation Timeline

Sources and References