NAIC Model Bulletin on the Use of AI Systems by Insurers
A text adopted in late 2023 by US insurance commissioners, taken up by many states, expecting insurers to maintain a written governance program for their AI systems and reminding them they are accountable for decisions made with them.
The Model Bulletin on the Use of Artificial Intelligence Systems by Insurers was adopted on December 4, 2023 by the US National Association of Insurance Commissioners. It creates no new rules: it restates that existing laws on unfair trade practices, rate discrimination and claims handling apply to decisions made with artificial intelligence, and it specifies what regulators expect in order to verify this. Each insurer must adopt a written governance program for its systems, proportionate to their use and impact on policyholders, covering risk management, internal controls, documentation, monitoring of drift and fairness of outcomes, and oversight of models and data acquired from third parties. The regulator can request access to that documentation during examinations. Taken up by many states, the bulletin brings US practice closer to an accountability logic similar to the European regulation, without adopting its risk classification. For the insurer, it means that a pricing or claims triage model bought from a vendor remains its responsibility, and that it must be able to explain to the regulator how it controls it.
Textbook case: a motor insurer uses a model bought from a vendor to flag claims suspected of fraud. During an examination, the state regulator asks for the model documentation, the fairness tests run on policyholders from different neighborhoods and the human review procedure for flagged files: the insurer must produce them, even though the model is not its own.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.