This bill would regulate the use of artificial intelligence in health insurance utilization review. It requires insurers to notify insureds, enrollees, and health care providers when AI-based algorithms are used in utilization review, to disclose certain information about those algorithms on request, and to identify AI use in adverse determination notices. The bill also requires insurers to submit AI algorithms and training data sets to the superintendent, conduct ongoing quality assurance testing, and certify that the tools minimize bias, follow evidence-based clinical guidelines, and do not independently create or change coverage criteria.
The bill further limits how AI may be used in coverage decisions. It provides that adverse determinations must be made only by clinical peer reviewers, that AI cannot be the sole basis for denying, delaying, or modifying medically necessary care, and that AI-based systems may not base adverse determinations solely on group data sets or directly or indirectly harm insureds or enrollees. It also expands adverse determination notice requirements to include detailed information about the reviewer’s credentials, time spent reviewing the case, and whether AI was used, and it updates the statutory definition of “emergency condition” in multiple insurance and public health law provisions to emphasize that the final diagnosis does not control the emergency determination.
The bill’s impact on state law would be significant for insurers, health maintenance organizations, and clinical peer reviewers operating under New York’s insurance and public health laws. It adds a new Insurance Law section governing AI in utilization review, amends existing notice provisions in Insurance Law and Public Health Law, and imposes compliance, reporting, audit, and penalty requirements enforced by the superintendent, with consultation from health and education officials as applicable. It also authorizes fines, license suspension or revocation, and other penalties for violations.
The general sentiment reflected by the bill text is strongly consumer-protective and oversight-oriented. The measure appears designed to increase transparency, ensure human clinical judgment remains central to utilization review, and address concerns about bias, opacity, and overreliance on automated decision-making in health coverage decisions. Because there are no committee transcripts or recorded votes provided, there is no direct evidence of support or opposition from debate or voting history.
The main points of contention likely concern insurer compliance burdens, disclosure of proprietary algorithms and training data, and the extent to which the state can require transparency over AI systems used in coverage determinations. Insurers may object to public reporting, audit access, and the requirement to submit algorithms and datasets to regulators, while patient advocates and providers would likely support the added disclosure, human-review safeguards, and anti-bias protections. Another possible area of debate is whether the bill goes far enough in restricting AI or whether it still permits meaningful use of AI in utilization review under regulatory supervision.
The bill would create a new regulatory framework in the Insurance Law for artificial intelligence used in utilization review and amend multiple Insurance Law and Public Health Law provisions governing adverse determinations and emergency condition definitions. It would require insurer disclosures, superintendent oversight, algorithm submission, quality assurance testing, and enhanced adverse determination notices, while also imposing penalties for noncompliance. The practical effect is to increase state oversight of insurer decision-making and to limit the role of AI in medical necessity determinations.
The bill’s overall tone is cautious and protective of patients, emphasizing transparency, human review, and anti-bias safeguards. With no committee transcript or vote record available, there is no documented legislative debate to indicate formal support or opposition, but the structure of the bill suggests a policy response to concerns about automated denials and opaque insurer practices.
Likely points of contention include the requirement that insurers disclose AI criteria, training data, and outputs; the mandate that AI not be the sole basis for adverse determinations; and the obligation to submit algorithms for regulatory review and publish quality assurance results. Insurers may view these provisions as burdensome or as exposing proprietary systems, while consumer advocates and health care providers are likely to support them as necessary protections against biased or unjustified coverage denials. The revised emergency-condition language may be less controversial, but it could still affect utilization review outcomes and appeals.