Connecticut sets five AI rules for state employee health plans, effective January 2027

Among them: no claim can be down-coded or a payment reduced by AI alone.

Comptroller Sean Scanlon announced five AI policies for the State Employee Health Plan and Connecticut Partnership Plan, covering more than 270,000 members. Among them: no adverse determination may be made solely by an AI system (human review required); carriers cannot use AI as the sole basis to downcode claims, reduce provider payments, or alter billing codes without human review; member data cannot be used to train other AI models. Effective January 1, 2027. CT Mirror independently reported the same announcement.

The source's own statement, as linked above. Nothing beyond it has been independently checked by us.

These are plan rules set by the Comptroller for plans the state sponsors; they are not a Department of Insurance regulation and do not reach commercial lines or other carriers' books.

For the health carriers and administrators serving those plans, AI in claims payment needs a human in the decision by January 2027. For everyone else it is a template to expect elsewhere.

• If you administer the state employee or Partnership Plan, confirm human review on any AI-assisted adverse claim action before January 1, 2027
• Treat this as a template, not a commercial-lines rule, unless the legislature extends it

Milliman launches an AI Solutions practice for insurance clients