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

What Is AI Insurance? Artificial Intelligence Liability Coverage Lines and Insurer AI Governance (NAIC & CMS Rules)

Technical guide to AI insurance across both enterprise domains: (1) insurance coverage for artificial intelligence systems (Tech E&O, cyber, medical malpractice, and AI model performance warranties) and (2) regulatory governance of AI in the insurance industry under the NAIC Model Bulletin and CMS 42 CFR § 422.101.

Written and maintained by CASRAI Editorial Board

Published

On this page

Enterprise risk managers, healthcare compliance officers, and AI governance teams searching for AI insurance (or artificial intelligence insurance) encounter two distinct operational domains that intersect in clinical informatics and commercial risk transfer:

  1. Insurance Coverage for Artificial Intelligence Systems: Commercial liability and performance policies that indemnify AI developers and enterprise deployers against algorithmic errors, hallucinations, bodily injury, intellectual property infringement, and adversarial attacks.
  2. Artificial Intelligence in the Insurance Industry: Predictive machine learning and generative AI models deployed by health payers and property/casualty carriers for underwriting, claims adjudication, and clinical utilization review—governed by the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers and federal CMS Medicare Advantage rules.

Understanding how traditional policy wordings leave silent gaps for autonomous model failures—and how state and federal insurance regulators audit algorithmic decision-making—is essential for hospital systems, clinical trial sponsors, and software vendors.

What Is AI Insurance?

AI insurance refers first to specialized commercial risk-transfer policies and endorsements designed to cover financial losses, third-party liability, and defense costs arising from the failure, miscalibration, hallucination, or regulatory breach of an artificial intelligence or machine learning model. It also encompasses the actuarial and regulatory framework governing how insurance carriers themselves deploy AI in underwriting and claims workflows.

Part 1: Insurance for Artificial Intelligence (Liability & Performance Coverage Lines)

When a generative AI application, clinical decision support (CDS) algorithm, or automated workflow agent produces an erroneous output, standard commercial policies often dispute coverage because traditional policy triggers distinguish between human negligence, network security breaches, and physical bodily harm. As analyzed in our comparison of AI liability insurance coverage lines, organizations must map AI failure modes across five distinct policy structures:

Insurance Coverage Line Core Insuring Agreement Covered AI Failure Modes Typical Exclusions & Silent AI Gaps
Technology Errors & Omissions (Tech E&O) Third-party pure financial loss resulting from a wrongful act, error, omission, or negligent rendering of technology services. Software bugs in AI pipelines, failure of an AI service to perform to contractual specifications, client financial loss from faulty algorithmic outputs. Excludes bodily injury and property damage (BI/PD); many legacy forms exclude intellectual property/copyright infringement from training data or intentional/unverified automated outputs.
Cyber Liability & Privacy Insurance First- and third-party losses from network security failures, unauthorized data access, and privacy statute violations. Prompt-injection data exfiltration, training data breaches exposing ePHI/PII, HIPAA Breach Notification costs, and regulatory fines where insurable. Does not cover model hallucinations, statistical drift, or algorithmic bias when no underlying network intrusion or security breach occurred.
Medical Professional Liability (MPL) / Healthcare Malpractice Bodily injury or wrongful death arising from professional medical rendering or failure to render healthcare services. Clinician reliance on an FDA-cleared Software as a Medical Device (SaMD) or ambient clinical scribe that contributes to diagnostic delay or dosing error. Disputes over product liability vs. professional negligence; unvalidated or unapproved AI tools used outside hospital governance policies may trigger policy conditions.
Dedicated AI Performance Warranties & Affirmative AI Endorsements Affirmative indemnification for contractual model accuracy degradation, hallucination liability, and third-party AI regulatory defense. Financial loss when a validated ML/LLM model falls below guaranteed statistical performance thresholds (e.g., sensitivity/specificity or error-rate bounds), copyright/IP defense, and EU AI Act / state bias audit defense. Requires rigorous pre-binding technical underwriting (model cards, out-of-sample calibration metrics, red-teaming logs, and drift monitoring).

How Underwriters Evaluate AI Systems Using NIST AI RMF 1.0

Specialty AI underwriters (such as Munich Re’s aiSure program and Lloyd’s-backed AI warranty syndicates like Armilla AI) do not price algorithmic risk from financial statements alone. Technical underwriting audits map directly to the four core functions of the NIST Artificial Intelligence Risk Management Framework (AI RMF 1.0, NIST AI 100-1) and the NIST Generative AI Profile (NIST AI 600-1):

  • GOVERN: Documented AI inventory, designated risk owners, human-in-the-loop escalation gates, and third-party foundation model vendor contracts.
  • MAP: Explicit intended-use boundaries, identification of high-stakes clinical or financial contexts, and data provenance documentation verifying training corpus licensing.
  • MEASURE: Quantitative pre-deployment benchmarks—including area under the receiver operating characteristic curve (AUROC), calibration slope, demographic subgroup parity, and adversarial prompt-injection pass/fail rates.
  • MANAGE: Production telemetry for covariate shift, automated kill-switches/fallback rules, and incident response playbooks.

Part 2: AI in the Insurance Industry (Underwriting & Payer Utilization Governance)

The second major dimension of AI and insurance governs how property/casualty, life, and health insurance carriers use predictive analytics and LLMs across the insurance lifecycle—and the binding regulatory rules enacted to prevent algorithmic discrimination and improper clinical denials.

1. The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers

On December 4, 2023, the National Association of Insurance Commissioners unanimously adopted the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, which has since been adopted by more than two dozen U.S. state insurance departments. As detailed in our comparison of insurer AI governance under the NAIC Model Bulletin, Colorado SB 21-169, and NY DFS Circular Letter No. 7, the Model Bulletin requires every licensed insurer using AI in underwriting, pricing, marketing, or claims to maintain a written Artificial Intelligence Systems (AIS) Program:

  • Proportional Risk Controls: Internal governance and validation rigor must scale with the degree of potential harm to consumers—with adverse underwriting declinations and claim denials subject to the highest scrutiny.
  • Bias and Proxy Variable Testing: Insurers must test predictive models before deployment and periodically thereafter to verify that external consumer data and information sources (ECDIS) do not act as unlawful proxies for protected classes under state Unfair Trade Practices Acts.
  • Third-Party AI Vendor Accountability: Insurers remain legally responsible for third-party actuarial, underwriting, and claims AI engines, requiring contractual audit rights and data lineage documentation capable of production during state Market Conduct Examinations.

2. Health Insurance AI: CMS Medicare Advantage Rules (42 CFR § 422.101)

In health insurance and clinical revenue cycle operations, payers increasingly use automated algorithms to screen prior authorization requests and inpatient/post-acute utilization management reviews. Following heightened regulatory scrutiny over algorithmic post-acute care denials, the Centers for Medicare & Medicaid Services (CMS) issued strict federal boundaries under CMS-4201-F (42 CFR § 422.101(c)) and the February 2024 CMS Health Plan FAQ:

  1. Individual Clinical Circumstances Mandate: While an algorithm or AI system may assist in predicting length of stay or flagging documentation elements, a Medicare Advantage organization must base coverage and medical necessity determinations on the individual patient’s specific clinical circumstances (medical history, physician progress notes, and individual functional status)—never solely on statistical averages from a larger dataset.
  2. No Shifting of Internal Coverage Criteria: Payer AI systems cannot apply internal, proprietary algorithmic thresholds that are more restrictive than publicly posted Traditional Medicare National Coverage Determinations (NCDs) or Local Coverage Determinations (LCDs).
  3. Mandatory Qualified Clinician Review on Adverse Determinations: Any adverse organization determination based on medical necessity must be reviewed and signed off by a physician or appropriate health care professional with expertise in the relevant field of medicine prior to issuing a denial.

Checklist for Enterprise AI Insurance & Governance Readiness

Before deploying a clinical or enterprise AI system—or renewing commercial liability policies—risk and compliance teams should complete five verification steps:

  • Audit Existing Tech E&O, Cyber, and MPL Policies for “Silent AI” Exclusions: Confirm in writing with your broker whether generative AI outputs, third-party LLM API calls, and automated decision support are affirmatively covered or excluded under current policy endorsements.
  • Maintain an Auditable Model Inventory & NIST AI RMF Dossier: Document every production model’s version, training/fine-tuning data provenance, out-of-sample validation metrics, and human-in-the-loop override logs to qualify for affirmative AI liability underwriting.
  • Enforce Human-in-the-Loop Attestation for Clinical & Coverage Decisions: Never allow an unreviewed model output to execute an autonomous medical diagnosis, treatment order, or payer coverage denial (42 CFR § 422.101(c)).
  • Establish Contractual Indemnification Allocations with AI Vendors: Verify whether foundation model providers and SaaS AI vendors provide IP copyright indemnification and how liability caps apply to data privacy breaches versus model output errors.

Frequently Asked Questions

What is AI insurance?

AI insurance refers both to specialized liability and performance insurance policies that protect organizations against financial losses caused by artificial intelligence systems (such as model hallucinations, algorithmic errors, copyright infringement, or clinical decision support failures) and to the use and regulation of AI models within the insurance industry for underwriting, pricing, and claims adjudication.

Does standard Tech E&O or Cyber insurance cover AI model hallucinations and errors?

Not always. Cyber insurance typically requires a network security breach or unauthorized data access trigger and does not cover financial losses caused solely by an accurate network delivering a hallucinated or statistically drifted AI output. Tech E&O covers third-party financial loss from technology errors but excludes bodily injury and may contain exclusions for generative AI intellectual property infringement or unverified automated outputs unless affirmative AI endorsements are added.

What rules govern how health and property insurers use AI in underwriting and claims?

State insurance regulators enforce the NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers (adopted December 2023) along with state statutes such as Colorado SB 21-169 and NY DFS Circular Letter No. 7, requiring a written AIS governance program, bias testing, and third-party vendor oversight. In healthcare, CMS regulates Medicare Advantage utilization management under 42 CFR § 422.101(c), prohibiting algorithmic coverage denials that bypass individual patient medical records and qualified clinician review.

Follow CASRAI

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

Ask CASRAI · free to try

Ask about What Is AI Insurance? Artificial Intelligence Liability Coverage Lines and Insurer AI Governance (NAIC & CMS Rules)

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 →