Skip to main content
v2026.11,986 entries · CC-BY 4.0

NAIC AI Bulletin (Model Bulletin on the Use of Artificial Intelligence Systems by Insurers): FACTS Principles, AIS Program Requirements, and State Adoption

A regulatory and compliance analysis of the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers (adopted December 4, 2023), detailing the FACTS principles, written AIS Program requirements, third-party vendor governance, and intersection with health insurance utilization management.

Written and maintained by CASRAI Editorial Board

Published

On this page

The NAIC AI Bulletin—formally titled the NAIC Model Bulletin: Use of Artificial Intelligence Systems by Insurers—was unanimously adopted by the National Association of Insurance Commissioners (NAIC) on December 4, 2023, following development by the NAIC Innovation, Cybersecurity, and Technology (H) Committee. The Model Bulletin establishes the primary U.S. regulatory framework governing how state-regulated insurance carriers (including health, life, property & casualty, and medical malpractice insurers) design, acquire, validate, deploy, and audit Predictive Models, Machine Learning (ML), and Generative AI systems across underwriting, pricing, marketing, utilization management, and claims adjudication.

Ask CASRAI · free to tryQuestion about this guide? Ask CASRAI answers with linked sources. 2 free questions a day, no account.

Ask your first question freeSubscribe

What Is the NAIC AI Bulletin? Statutory Authority and the FACTS Principles

Rather than creating a brand-new insurance statute from scratch, the December 2023 NAIC Model Bulletin clarifies how existing state insurance laws apply to algorithmic decision-making and operationalizes the five NAIC Principles on Artificial Intelligence (the “FACTS” Principles) originally adopted by the NAIC in August 2020:

  • F — Fair and Ethical: Insurers must take proactive steps to avoid proxy discrimination and unfair bias against protected classes across the AI lifecycle, ensuring outcomes comply with state insurance laws.
  • A — Accountable: Insurers are legally responsible for the compliance and outcomes of all AI systems they deploy—even when the model or training dataset is licensed from an external third-party software vendor.
  • C — Compliant: AI systems must conform to all applicable federal and state insurance statutes, specifically the Unfair Trade Practices Model Act (Model #880), the Unfair Claims Settlement Practices Model Act (Model #900), the Unfair Life, Accident and Health Claims Settlement Practices Model Regulation (Model #902), and the Corporate Governance Annual Disclosure Model Act (Model #305).
  • T — Transparent: Insurers must maintain documentation enabling both state insurance regulators and impacted consumers/providers to understand how AI systems influence adverse insurance decisions.
  • S — Secure, Safe, and Robust: AI systems must be protected against cybersecurity threats, adversarial manipulation, model drift, and unintended operational failures.

In Section 2 of the Model Bulletin, the NAIC defines an “Adverse Consumer Outcome” as a decision made by an insurer using an AI System that results in an outcome adverse to the consumer (or policyholder/beneficiary), such as denial, cancellation, or non-renewal of coverage, an increase in premium rate, or the reduction, delay, or denial of a claim or prior authorization request.

Structure of the NAIC Model Bulletin: The Written AIS Program (Section 3)

The core operational mandate of the NAIC Model Bulletin (Section 3) is that every licensed insurer using AI systems must develop, implement, and maintain a written Artificial Intelligence Systems Program (AIS Program). The bulletin emphasizes proportionality: the rigor of controls in an insurer’s AIS Program must scale with the degree of potential harm to consumers, whether the AI output is augmenting a human reviewer or fully automating an Adverse Consumer Outcome, and the complexity/opacity of the model.

AIS Program Pillar (NAIC Section 3) Mandatory Insurer Governance & Technical Controls Documentation Expected in Market Conduct Exams (Section 4)
1. Governance Framework Board of directors or senior executive accountability; cross-functional AI governance committee uniting actuarial, clinical/medical directors, legal, compliance, data science, and IT security; clearly defined escalation chains and human-in-the-loop authority. Board minutes, committee charters, written AIS Program policies, and organizational role matrices.
2. Risk Management & Internal Controls Centralized enterprise AI System Inventory tracking all predictive and generative models across marketing, insurance underwriting, rating, utilization management, fraud detection (SIU), and claims; formal change-management and validation gates before production release. Complete AI inventory register, model risk tiering classification, and pre-deployment sign-off packets.
3. Data & Model Lifecycle Validation Evaluation of training/input datasets for completeness, accuracy, and unfair demographic bias; statistical testing for unfair discrimination (under Model #880); and continuous post-deployment monitoring for model drift (covariate shift and concept drift). Data provenance dictionaries, bias/disparate-impact test protocols, validation metrics, drift thresholds, and remediation logs.
4. Third-Party AI & Data Vendor Oversight Formal pre-acquisition due diligence on third-party AI vendors and external consumer data sources; contractual clauses requiring vendor cooperation with state insurance department audits, notification of material model updates, and right-to-audit provisions. Third-party vendor due diligence questionnaires, procurement contracts, audit reports, and vendor model cards.

How Is the NAIC AI Bulletin Used in Healthcare and Health Insurance Governance?

In healthcare delivery, clinical research, and health plan administration, the NAIC AI Bulletin intersects directly with federal Centers for Medicare & Medicaid Services (CMS) rules and state-level insurer AI governance frameworks (such as Colorado SB 21-169 and New York DFS Circular Letter No. 7 of 2024):

  1. Prior Authorization and Utilization Management (UM) Algorithms: Health insurers increasingly deploy predictive algorithms and natural language processing (NLP) to screen prior authorization requests, predict length of stay, and flag post-acute care extensions. Under the NAIC Model Bulletin—and parallel federal Medicare Advantage rules at 42 CFR § 422.101(c) (CMS-4201-F)—an AI system cannot be used as the sole basis to deny or terminate medically necessary care without individual clinical review by a qualified healthcare professional against publicly accessible clinical criteria.
  2. Prohibition on Non-Actuarial Proxy Discrimination: When health, disability, long-term care, or supplemental insurers ingest external consumer data or Social Determinants of Health (SDOH) variables into predictive models, the AIS Program must verify that those features have a demonstrable actuarial nexus to risk and do not act as unlawful proxies for race, ethnicity, sex, or protected health status under the Unfair Trade Practices Model Act (`#880`).
  3. Third-Party Vendor Liability Cannot Be Outsourced: Many health plans license commercial utilization management platforms, payment integrity engines, or Special Investigations Unit (SIU) fraud-scoring algorithms from third-party vendors. Section 3 of the NAIC Model Bulletin establishes that the licensed insurer remains 100% regulatory accountable for vendor algorithms and must secure contractual audit rights so state insurance examiners can inspect how vendor models operate. See also AI Insurance and Insurer AI Governance and AI Governance in Healthcare.

State Adoption of the NAIC AI Bulletin and Market Conduct Examinations

Because the NAIC is a standard-setting organization of state insurance commissioners rather than a federal legislature, the Model Bulletin takes binding supervisory effect in each U.S. jurisdiction as individual State Departments of Insurance (DOIs) issue it as an official state insurance bulletin, circular letter, or administrative regulation:

  • Broad Multi-State Adoption (2024–2026): Following its December 4, 2023 adoption by the NAIC Plenary, more than 20 U.S. jurisdictions—including Alaska, Connecticut, District of Columbia, Illinois, Kentucky, Maryland, Michigan, Nebraska, Nevada, New Hampshire, Pennsylvania, Rhode Island, Vermont, Virginia, Washington, and West Virginia—have formally adopted the NAIC Model Bulletin (with states such as California, Colorado, and New York enforcing parallel or stricter state-specific AI insurance circulars and statutes).
  • What Examiners Request Under Section 4: During routine or targeted Market Conduct Examinations under the NAIC Market Regulation Handbook, state examiners can compel an insurer to produce: (1) its written AIS Program documentation, (2) its inventory of all internal and vendor AI systems affecting underwriting or claims, (3) empirical evidence of bias and drift testing, and (4) third-party vendor contracts and validation artifacts.

Frequently Asked Questions

What is the NAIC AI Bulletin?

The NAIC AI Bulletin (Model Bulletin on the Use of Artificial Intelligence Systems by Insurers), adopted on December 4, 2023 by the National Association of Insurance Commissioners, is the foundational regulatory framework directing state-licensed insurers to establish a written Artificial Intelligence Systems (AIS) Program. Grounded in the NAIC’s five FACTS principles and the Unfair Trade Practices (`#880`) and Unfair Claims Settlement Practices (`#900`) Model Acts, it governs AI risk management, bias and drift testing, human oversight, and third-party vendor auditing.

How is the NAIC AI Bulletin used in clinical research and healthcare?

In healthcare and health insurance, state insurance commissioners use the NAIC AI Bulletin to regulate predictive and generative AI models deployed by health plans, medical malpractice carriers, and third-party administrators (TPAs) for prior authorization, utilization management, claims adjudication, and actuarial underwriting—requiring documented clinical oversight, validation against unfair discrimination, and contractual auditability of third-party healthcare AI vendors.

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about NAIC AI Bulletin (Model Bulletin on the Use of Artificial Intelligence Systems by Insurers): FACTS Principles, AIS Program Requirements, and State Adoption

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →