For Insurers

Evidence for a bounded reliance decision

HEART can organize governance evidence relevant to underwriting, but it does not establish that a control, score, credential, or professional title predicts insured loss. Predictive value requires claims and loss data, valid comparators, confounder analysis, calibration, and independent review.

Underwriting questions HEART can structure

Question	Evidence focus
What is insured?	Exact system, function, deployment, population, jurisdiction, period, and exclusions.
Which duties apply?	Current law, contract, policy, standards, and Profile sources.
Do controls operate?	Requirement-level evidence, execution records, tests, change history, incidents, and remedy.
What could invalidate reliance?	Model and deployment change, stale evidence, scope drift, status transitions, dependency failure, or source change.
Who made the finding?	Competence, independence, method, authority, conflict, review, and decision records.
What does a credential mean now?	HVC scope and limitations plus current Registry status and lineage.

No unvalidated tier pricing

HVC v4.0 does not use Gold, Silver, Bronze, or a universal BGF score. HEART does not claim that an HVC, Guardian assessment, GTE, MAP-States trace, Behavioral Oracle attestation, or Trust Infrastructure Index corresponds to a premium, deductible, coverage, reserve, or loss tier.

A carrier may research those relationships under the current v2.0 interface, but should preregister outcomes where appropriate, protect claims data, account for selection and reporting bias, retain null findings, and avoid circular validation in which underwriting treatment creates the measured outcome.

Insurers’ own use of AI and underwriting another organization’s AI-related risk are separate contexts. NAIC AI governance materials principally address insurer conduct; they do not establish insurer acceptance of HEART as an underwriting factor.

See Assurance, Methods, and For Researchers.