October 1 Meeting Summary (from the Wheeler Report)

Regarding consumer-facing chatbots, members leaned toward requiring upfront disclosure rather than regulating how chatbots behave. They still asked staff to include draft [legislative] options for suicide and self-harm protocols and for extra safeguards when a chatbot interacts with a minor, so those can be debated and possibly cut later. The committee chose not to require healthcare practitioners to disclose when they use AI tools. Members said AI is already embedded throughout clinical work, citing radiology as an example, and that disclosure rules would become a "slippery slope." This was consistent with the Medical Examining Board's position that the practitioner is responsible for the record. That also produced a discussion over ambient-listening scribes that delete their audio recordings, and whether patients lose evidence they could use in a malpractice dispute. Members likewise declined to mandate AI continuing education or to legislate governance standards for organizations that deploy AI. With regard to governance, several members said smaller and rural providers need resources such as grants and data-science support more than they need new requirements.

On health insurance and prior authorization, there was broad support for requiring meaningful human review before a coverage or prior authorization request is denied, along with reporting on outcomes so the requirement can be enforced. Another topic was a state AI "regulatory sandbox," which members suggested renaming a "laboratory." It would grant AI developers temporary, supervised relief from state rules, mainly licensing rules, so they could test products in Wisconsin and report the results back to the state. Members discussed which agency would run it, how much it would cost, whether there is real demand, how patients would be protected, and how it could connect to a multi-state effort.