Connecticut Restricts AI From Making Health Coverage Decisions Alone

HARTFORD, Connecticut — Artificial intelligence will not be allowed to make certain health coverage decisions without human review for more than 270,000 people enrolled in Connecticut state health plans.

State Comptroller Sean Scanlon announced new protections governing how insurance companies and healthcare administrators use artificial intelligence in the State Employee Health Plan and Partnership Plan.

"You and your doctor should make your health care decisions, not artificial intelligence."

The new policy establishes five principal protections.

Human Review of Healthcare Decisions

An insurance company cannot use artificial intelligence as the only basis for denying coverage or making another decision that could negatively affect a patient.

A human must review adverse determinations involving healthcare services or benefits.

Protections for Medical Providers

Insurance carriers cannot rely solely on AI to reduce provider payments or change medical billing codes.

Authorized automated claim processing can continue, but algorithms cannot replace human review when a decision may negatively affect a doctor, hospital or other healthcare provider.

Patient Privacy

Information belonging to members of the state health plans cannot be used to train, develop or support unrelated artificial intelligence models.

Medical information may include diagnoses, medications, treatments, claims and other personal details.

Disclosure When AI Is Used

Patients must be informed when artificial intelligence is materially assisting with a service, communicating directly with them or recommending a health benefit or service.

A patient should not be led to believe that an automated system is a human representative.

Accuracy and Fairness

Insurance carriers must examine their AI systems for accuracy, consistency and fairness.

Companies will also be required to disclose their AI governance and auditing procedures to the comptroller's office.

The policy is intended to prevent automated systems from repeating errors or inequalities that may already exist in healthcare information.

Who Is Protected?

The protections currently apply to more than 270,000 members of Connecticut's State Employee Health Plan and Partnership Plan.

They do not yet cover every Connecticut resident or every private health insurance plan.

Scanlon said he plans to work with the Connecticut General Assembly during the 2027 legislative session to seek similar protections for all state regulated health plans.

The policy does not prohibit the use of artificial intelligence in healthcare. AI may still assist with administrative work and authorized claims processing.

The central requirement is that artificial intelligence cannot act as the only decision maker when a determination could negatively affect a patient or medical provider.