Senate panel advances bill requiring human review and disclosure when AI influences coverage decisions

The Senate Committee on Insurance advanced legislation on March 11 that would require human clinical review and greater disclosure when insurers use artificial intelligence or automated decision systems in making coverage determinations.

Senator Luneau, sponsor of Senate Bill 246, told the committee the bill "allows for the human review of artificial intelligence when it's used in a patient setting" and gives patients the right to appeal adverse decisions that relied on such systems.

The committee adopted a series of technical and substantive amendments, including a set offered by Senator Bass that defines an "automated decision system" to cover processes "derived from machine learning, statistical modeling, data analytics, or artificial intelligence that issues simplified output, including but not limited to a score, classification, or recommendation that is used to assist or replace human discretionary decision making and materially impacts a natural person." Senator Luneau said the definition was written broadly to avoid loopholes.

Key provisions added in the amendment set require:
- A licensed physician to sign any adverse determination and certify they personally reviewed the medical record and bear responsibility for the clinical judgment; Senator Bass described that requirement as intended to prevent a "rubber stamp" by an algorithm.
- Disclosure to the enrollee and the Department of Insurance when an AI or automated decision system was used in any part of the coverage determination, and documentation of the extent to which the system influenced the decision.
- A presumption that an adverse determination materially contributed to by an automated system is invalid unless the insurer demonstrates the final determination was independently reached through documented clinical judgment.
- Authority for the insurance commissioner to inspect and audit any AI or automated decision system used in utilization review, request submission and independent review of systems, and review related policies and procedures; upon request insurers must disclose data sources, training parameters and validation methods and pay for any independent review the commissioner deems necessary.

Senator Barrow asked how and when enrollees would be notified that AI was used in a decision and urged clarity on the notification procedure. Senators said the Department of Insurance would adopt rules to implement statutory requirements and that the sponsors proposed a start date of Jan. 1, 2027, to allow time for regulators and insurers to prepare. The committee also adopted language that would require policies in force before that date to conform upon renewal but no later than Jan. 1, 2028.

Committee members discussed safeguards if AI recommendations conflict with clinical review. Peter Robbins Brown of Louisiana Progress said the bill's requirement for a meaningful human review was intended to make clinical judgment the prevailing decision. A committee member read bill language that "if an adverse determination is appealed on the basis of the use of an artificial intelligence, the insurer shall not use artificial intelligence in any subsequent review of the claim."

The committee noted witness cards filed in support and, after debate, the chair moved to report SB246 as amended. Chair said there was no opposition and the bill was reported out of committee as amended. No recorded roll-call vote was taken; adoption and reporting were announced by voice.

The committee did not set a floor vote date during the meeting; sponsors said they will work with stakeholders and with the Department of Insurance on rule details and implementation timing.