Health facilities and plans would disclose when they use AI

Facilities would have to tell patients when they use AI for clinical decisions and follow rules for its use. This covers the health care settings listed in the bill, including hospitals, clinics, and testing facilities. AI could not replace a provider’s clinical judgment, discriminate unlawfully, or create foreseeable, material risks of harm to patients. AI-generated messages about a patient’s clinical information must say they were made by AI and explain how to contact a person. This does not apply to administrative messages or messages a provider individually reviewed. Facilities using AI for clinical decisions would have to file a yearly statement describing the systems and how they oversee them. The bill would take effect one year after enactment.

Health insurers would have to disclose AI use in coverage reviews and have providers review certain benefit decisions. This covers health insurers and providers who conduct coverage reviews for them, and people covered by the insurers’ health policies. AI decisions must consider the person’s clinical history, individual circumstances, and other relevant records; they could not rely only on group data. Before denying, reducing, or ending benefits, including denying a prior authorization request, a provider conducting the review must check and document individual records and use independent judgment. Insurers using AI in coverage reviews would have to file a yearly statement describing the systems and how they oversee them. The bill would take effect one year after enactment.

MA and CHIP managed care plans would have to disclose AI use in coverage reviews and have providers review certain benefit decisions. This covers MA and CHIP managed care plans and providers who conduct coverage reviews for them, and people enrolled in those plans. AI decisions must consider the enrollee’s clinical history, individual circumstances, and other relevant records; they could not rely only on group data. Before denying, reducing, or ending benefits, including denying a prior authorization request, a provider conducting the review must check and document individual records and use independent judgment. Plans using AI in coverage reviews would have to file a yearly statement describing the systems and how they oversee them. The bill would take effect one year after enactment.