Health Occupations - Behavioral Health Care Providers - Use of Artificial Intelligence
HB995 would regulate the use of artificial intelligence by behavioral health care providers in Maryland. The bill generally prohibits licensed, registered, or certified behavioral health professionals from using AI to provide behavioral health care directly to a patient, including for assessment, diagnosis, treatment, counseling, treatment planning, case management, and therapeutic communications. It defines key terms such as behavioral health care, behavioral health care provider, administrative support tasks, and therapeutic communication.
The bill does allow AI systems to be used for limited administrative support functions, such as scheduling, records and billing management, data analysis for operational purposes, and organizing session notes, but only if the provider complies with confidentiality and health-record security laws, signs an agreement preventing the use of patient information to train AI systems, and gives patients written and verbal notice about AI use and its risks. If AI is used in a way that could materially affect clinical decision-making or patient-facing services, the provider must obtain written informed consent. The bill also requires providers to independently review AI-generated reports or summaries and preserves the ability to distribute self-help or educational materials that do not claim to provide behavioral health care.
HB995 would add a new section to the Health Occupations Article, creating enforceable standards for behavioral health professionals’ use of AI and making violations subject to disciplinary action by the appropriate licensing board. It would not ban AI in all clinical settings, but it would sharply limit AI’s role in direct patient care and impose privacy, disclosure, consent, and oversight requirements when AI is used for administrative or potentially patient-facing functions. The bill would affect licensed, registered, and certified behavioral health providers whose scope includes behavioral health services, as well as vendors or system owners that contract with those providers.
Based on the bill text and the absence of recorded committee testimony or votes, the overall posture of the bill appears precautionary and protective rather than permissive. The measure reflects concern about patient safety, confidentiality, and the limits of AI in mental and behavioral health treatment. Its structure suggests support for narrow administrative uses of AI while rejecting AI as a substitute for professional judgment in therapeutic care.
The main point of contention is likely the bill’s broad prohibition on AI for direct behavioral health care, which could be viewed by providers and technology vendors as overly restrictive or difficult to apply in modern practice. Another likely issue is the informed-consent and notice requirements, especially the obligation to warn patients about confidentiality risks and to secure agreements preventing training on patient data. Supporters would likely emphasize privacy, accuracy, and the need for human oversight in sensitive mental health services, while opponents may argue the bill could limit innovation, increase administrative burden, and create uncertainty about what counts as AI-assisted clinical decision-making or patient-facing services.