Valued Voice

The Legislative Council Study Committee on the Use of Artificial Intelligence (AI) in Health Care met Nov. 12 to review the preliminary drafts of legislation developed from its discussions earlier this fall. Ahead of the meeting, WHA submitted a memo to committee members with comments, questions and concerns regarding the drafts.

As covered previously in The Valued Voice, the committee is led by Sen. Rachael Cabral-Guevara and is made up of four bipartisan state legislators and 12 public members, including four leaders from Wisconsin health systems. Since its first meeting in August, the committee has heard from health care organizations, national policy experts, regulators from other states, payers and providers. During its October meeting, committee members discussed policy options and asked committee attorneys to prepare draft legislation on several topics, including:

• Consumer data privacy
• AI systems representing themselves as licensed professionals
• Human review of insurance coverage decisions
• Companion chatbots, and
• An “AI regulatory sandbox”

WHA’s memo builds on its September recommendations to the committee, reiterating that AI use by hospitals and health systems is not occurring in a regulatory vacuum and reminded the committee that our members must already comply with numerous federal and state laws including HIPAA, the HITECH Act, the 21st Century Cures Act, comprehensive health information privacy statutes in state law, professional licensure and more.

Therefore, WHA suggested the committee focus on regulatory gaps created by consumer grade AI tools that may handle identifiable health information with little to no accountability. In several areas, WHA encouraged the committee to address those gaps with Wisconsin’s existing laws rather than creating new statutory frameworks.

On data privacy, for example, WHA agreed that consumer AI tools not connected to licensed health care providers, such as ChatGPT Health and programs connected to wearable devices, likely fall outside of existing privacy laws. However, rather than creating a new regulatory scheme that may also end up impacting health care providers, WHA’s comments recommended updating the definition of “health care provider” in Wisconsin’s long-standing health care confidentiality statute to include consumer AI health tools. That would make those systems subject to the same confidentiality requirements as established providers. WHA also raised concerns about a bill draft that would make certain uses of “sensitive data” a felony, particularly because it includes no intent requirements and no exemption for HIPAA-covered entities.

Regarding professional licensure, WHA supported the concept of prohibiting an AI tool from being licensed or from representing itself as a licensed health care professional. However, the memo noted, for example, that state law already prohibits any “person” from practicing medicine without a license, and therefore suggested clarifying this prohibition could be updated to also apply to AI tools. WHA also again cautioned against penalties written into the early draft that may impose criminal liability on providers that deploy AI, rather than on the developers best positioned to control the AI’s underlying programming. WHA noted this issue could raise serious legal questions about clinical use of AI tools in caring for patients.

WHA expressed support for a bill draft addressing AI in health insurance coverage decision making. The bill drafted for the committee would prohibit insurers from issuing denials based solely on an AI tool’s determination without meaningful human review. It would also require annual reporting of adverse determinations recommended by AI. In the memo, WHA emphasized that patients, families, health care providers and hospitals all feel the effects of those decisions. The memo suggested two additions: (1) requiring the human reviewer to have “appropriate clinical expertise” and (2) broadening “adverse determination” to include partial denials and payment reductions, not just full denials.

WHA also supported the committee’s intent of addressing risks from consumer companion chatbots, but questioned how the bill draft would apply to tools used by health care providers within a treatment relationship.

Finally, WHA’s memo supported the concept behind a so-called AI “regulatory sandbox” in Wisconsin, as noted earlier in the Association’s September memo. As an alternative, WHA suggested the legislation could instead give certain state agencies authority to grant temporary waivers for innovative uses of AI.

Wisconsin’s Legislative Council Study Committee is scheduled to meet again in December and will then make recommendations to the Legislature. WHA will continue to engage with the study committee as it evaluates potential policy recommendations affecting hospitals, health systems, clinicians and patients. 

For more information, contact WHA Policy Counsel Andrew Hanus.