Senate amendment would bar companies from offering AI mental‑health services without clinician oversight

The House committee reviewed the Senates amended language for bill 8163.1, which would restrict use of artificial intelligence to provide mental‑health services unless a licensed clinician reviews and approves those services.

"It is the purpose of this act to safeguard individuals seeking mental health services in Vermont from psychological harm, including death by suicide, by ensuring that these services are delivered by mental health professionals and not independently by artificial intelligence systems," said Katie McLean of the Office of Legislative Counsel as she walked members through the Senate Health and Welfare committees changes.

The amendment defines "mental health services" to include counseling, therapy or psychotherapy used to diagnose or treat mental or behavioral health conditions, provide recovery support, make therapeutic decisions, issue direct therapeutic communications, generate treatment plans or recommendations, or detect or interpret emotion or mental states. Under the committees markup, an individual clinician may use HIPAA‑compliant AI tools provided the clinician reviews and accepts any AI output before applying it with a client. By contrast, subsection b was rewritten to prohibit corporations or entities from providing, advertising, or otherwise offering mental‑health services via AI to the public unless those services are provided by a mental‑health professional.

McLean said the draft the committee saw reflects committee markup, not a subsequent floor amendment. "There was a floor amendment to remove individual on subsection b and c," she said, noting the floor change could narrow the bills reach to companies and entities rather than individual users.

Enforcement in the draft would include civil penalties and Consumer Protection Act liability. McLean told the committee that violations by a corporation or by an unlicensed person "shall be subject to the Consumer Protection Act and that each violation shall carry a civil penalty of $10,000." The text also adds a reporting requirement directing the Office of Professional Regulation and the Board of Medical Practice to submit recommendations next January on the regulation of AI by regulated professionals.

Members raised definitional and public‑safety questions. One lawmaker warned that an overly broad inclusion of "detecting or interpreting emotion or mental states" could sweep in ordinary, nonclinical uses and impede tools that currently direct distressed users to hotlines. "If we prohibit it from being able to identify if someone's suicidal or depressed or anxious, then we might actually contribute to deaths because it won't catch them," a committee member said, urging caution about restricting life‑saving detection features.

Counsel and members debated whether to align Vermonts definition with other states; Maines law was cited as having a different, shorter definition that excludes "therapeutic decisions." McLean noted those are policy choices the committee must make.

Action and next steps: the committee chair said she was not prepared to support the amended Senate version and asked that the item be delayed for additional testimony. Members agreed to seek more information before concurring or making further amendments. No formal vote was recorded during the session.

What happens next: committee staff will work with counsel to clarify definitions and exceptions, gather additional testimony, and return with revised language before the committee considers formal action.