NAIC AI Model Bulletin
This section is the “why now” of the Glacis strategy, grounded in the Law360 article “Tracking the Evolution of AI Insurance Regulation in 2025” by Heidi Lawson, Yanfei Wang & Sarah Hopkins (Fenwick & West LLP), Dec 11 2025. Heidi Lawson is the same “Heidi” who outlined the Glacis revenue streams — so this article is effectively the regulatory thesis behind the strategy. GROUNDED · Law360
The big picture
Section titled “The big picture”AI now affects underwriting, pricing, claims processing, and customer engagement. 2025’s headline development was agentic AI — autonomous systems performing insurance tasks without human input — which pushed the NAIC to escalate oversight so innovation doesn’t outpace consumer protection. GROUNDED · Law360
The central instrument: the AI Model Bulletin
Section titled “The central instrument: the AI Model Bulletin”The NAIC’s governance instrument is the December 2023 “Model Bulletin: Use of Artificial Intelligence Systems by Insurers.” Its defining quality: it is principle-based, not prescriptive. GROUNDED · Law360
- It requires insurers to establish governance, documentation, and audit procedures —
- but it does not prescribe specific standards.
- Enforcement leans on EXISTING law: unfair trade practice statutes, unfair claim settlement practice statutes, and consumer protection laws — not a dedicated AI statute. GROUNDED · Law360
This is the crucial cross-link to your exam study: the enforcement backbone for AI is the same Unfair Trade Practices Act and Unfair Claims Settlement Practices Act you learn in 12 Ethics & trade practices. The “new” AI rules ride on old, well-tested consumer-protection law.
Adoption (a moving number — VERIFY)
Section titled “Adoption (a moving number — VERIFY)”By late 2025, 23 states + Washington, D.C. had adopted the NAIC AI Model Bulletin, with some variations. GROUNDED · Law360
The survey findings (VERIFY the percentages)
Section titled “The survey findings (VERIFY the percentages)”The NAIC Big Data and AI (H) Working Group’s 2025 health-insurance survey (building on prior auto/life/home studies) reported AI adoption: 92% of health insurers, 88% auto, 70% home, 58% life report current or planned AI use. Yet nearly one-third of health insurers still don’t regularly test their models for bias/discrimination — exactly the gap regulators fear. GROUNDED · Law360
What a compliant AI insurer must demonstrate (stable principles)
Section titled “What a compliant AI insurer must demonstrate (stable principles)”Strip away the volatile counts and the durable expectations are clear — and they map directly onto Glacis’s product:
- Governance over AI use,
- Documentation of models and decisions,
- Audit procedures,
- Bias/discrimination testing,
- Human oversight (especially as agentic AI removes humans from the loop),
- (emerging) consumer disclosure when AI drives a decision.
This is the “compliance-as-a-product” thesis: an AI-native platform that ships these as features turns regulation from a tax into a moat. GROUNDED · Glacis
Why it matters for all three Glacis streams
Section titled “Why it matters for all three Glacis streams”- Underwriting SaaS — the bulletin’s governance/documentation/audit asks become a product-requirements spec.
- Producer / MGA (healthcare Tech E&O) — healthcare is the highest-adoption, highest-litigation line, so a healthcare-AI MGA must be governance-native.
- TPA / claims attribution — claims AI is the hottest litigation zone; attribution maps to the documentation/explainability demands.
→ 01 AI evaluation tool & third-party WG
Drill this page →5 bank questions stand behind what you just read. Check it while it’s warm.