States are increasingly allowing artificial intelligence to assist with healthcare while drawing a line at letting it independently deny coverage, provide mental health treatment or interact with patients without their knowledge.
Lawmakers in 11 states advanced 14 healthcare-related AI policy actions during the first half of 2026, according to the Transparency Coalition’s Mid-Year State AI Legislation Report. The measures were part of 84 AI-related bills passed or enacted in 27 states.
The largest group addresses insurer use of AI in prior authorization, coverage and payment decisions. Alabama, Colorado, Georgia, Illinois, Iowa, Utah and Washington adopted measures requiring varying degrees of human oversight, individualized review or disclosure.
See also: AI’s real hurdle in healthcare? It’s all in our heads
State regulations aim to prevent bad healthcare decisions based solely on an algorithm
The laws generally stop short of prohibiting AI. Instead, they seek to prevent adverse decisions from being based solely on an algorithm or population-level data without consideration of a patient’s medical history and clinical circumstances.
Alabama requires insurers to disclose their use of AI and reserves denials for licensed health care professionals. Colorado requires a licensed clinician to review coverage denials, while Washington prohibits AI from serving as the sole means of denying, delaying or modifying healthcare services. Illinois’ law applies to automated claim downcoding and requires those determinations to be made or reviewed by a person.
The restrictions arrive amid scrutiny of prior authorization and insurers’ use of predictive technology. Federal investigators and lawmakers have raised concerns about high Medicare Advantage denial rates and the possibility that employees may be pressured to follow machine-generated recommendations. Insurers have maintained that algorithms do not make final care decisions.
Mental health chatbots prompted another cluster of laws. Colorado, Maine, Rhode Island, Tennessee and Vermont restricted AI systems from independently providing therapy, making treatment decisions or presenting themselves as qualified mental health professionals.
Mental health experts have warned that general-purpose chatbots may provide inaccurate advice, reinforce harmful beliefs, encourage emotional dependence or fail to recognize when a user needs emergency intervention. The state laws vary in scope. Tennessee prohibits developers and organizations from representing that an AI system can act as a qualified mental health professional, while Rhode Island prohibits licensed providers from allowing AI to make independent therapeutic decisions or determine treatment plans.
Other measures focus on transparency. Iowa requires health care professionals to disclose the use of recording technology before capturing a patient encounter for AI transcription. Utah clarified that technology providing advice or treatment without interaction between a patient and practitioner does not qualify as a healthcare innovation for scope-of-practice purposes.
The compliance challenge will become more immediate in 2027. Although several measures have already taken effect, insurer-related laws in Colorado, Georgia, Illinois and Utah, along with Rhode Island’s mental health AI restrictions, are scheduled to take effect Jan. 1.
Credit: Source link
