Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Artificial Intelligence topic
No spam. Unsubscribe anytime.
Regulators and board wrestle with fast‑moving AI in health care after FSMB briefing
Summary
Frank Myers of the Federation of State Medical Boards told the Medical Board of California that AI is expanding rapidly in clinical documentation, triage and direct‑to‑consumer services and urged state boards to focus on education, physician accountability, consent/privacy, bias mitigation and human‑in‑the‑loop safeguards.
Frank Myers, director of regulatory innovation at the Federation of State Medical Boards, urged the Medical Board of California on Thursday to prepare for rapid changes in health‑care artificial intelligence while keeping patients safe.
Myers told the board that large language models and ambient listening tools are already being used to generate clinical notes, assist triage and, increasingly, provide direct‑to‑consumer health guidance. He singled out ambient‑listening systems that transcribe entire patient encounters and then create notes — systems that one slide he cited said had been deployed by more than 150 health systems — and warned regulators to pay attention to data retention, consent and transcription errors.
“Who is responsible when an AI misinterprets an audio recording and populates the record with information that isn’t relevant?” Myers asked. He described a four‑part risk framework for regulators: poor implementation, failure to define professional duties, deliberate bad actors and hallucinations (confident but incorrect AI outputs).
Myers reviewed the current regulatory patchwork — with many states passing different AI rules — and described recent California laws that he said had moved the state forward: protections for professional titles, training‑data disclosures, and requirements for disclosure when AI generates communications with patients. He also described a high‑profile Utah pilot, in which a company called Doctronic provided AI‑driven prescription renewals under a 12‑month mitigation waiver; that pilot uses a phased human‑review model but raised concerns about statewide deployment prior to full oversight.
Board members asked whether recordings and transcripts used by AI become part of the legal medical record and how long audio files should be retained. Myers advised regulators to consider subpoenas for audio in investigations, and he recommended that physician education and informed consent be central parts of oversight.
Public commenters — including the California Medical Association — said they supported strong oversight and transparency. Some public speakers pressed for more clarity about how the board would respond to specific risks, echoing concerns about direct‑to‑consumer AI that gives medical‑style advice without clinician oversight.
The board did not take immediate regulatory action but members signaled interest in making AI a recurring agenda item and in coordinating with other state agencies to clarify rules on disclosure, consent and medical‑record content.
Next steps: board members requested more guidance on whether audio recordings should be treated as part of the medical record, and staff said they would explore interagency collaboration and follow up with resources Myers cited.
