Skip to content

mizcausevic-dev/unfair-discrimination-incident-card-profile

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

unfair-discrimination-incident-card-profile

Unfair Discrimination AI Incident Card Profile v0.1 draft. Profile of the AI Incident Card spec scoped to insurance unfair-discrimination, biased decisioning, NAIC AI Model Bulletin governance gaps, ECDIS input defects, FCRA-dispute-pattern AI events, and state DOI exam findings. Provides the severity / event_type / regulator-referral taxonomy a state DOI market-conduct examiner + Fair Insurance Committee + actuarial committee + outside counsel expect when an AI insurance tool is implicated in a consumer harm.

Part of the Kinetic Gain Protocol Suite.

Status: v0.1 draft. Profile at profile.json, canonical example at examples/coastguard-aian-rural-incident.json.

Why this exists

When an AI insurance tool is implicated in a consumer harm — a patterned four-fifths violation, an ECDIS defect, an FCRA dispute pattern, an actuarial-soundness gap, a third-party vendor control failure, or a state DOI exam finding — the insurer's compliance / actuarial / outside-counsel / Fair Insurance Committee chain needs a single canonical record covering: what happened, who was affected, what AI system was involved, what state's regulators have jurisdiction, what referral pathways apply, what's the remediation plan, and is it signed. State DOI examiners triangulate from this record to the underlying audit-stream events, bias-coverage bundle, vendor due-diligence package, and Decision Card.

This profile defines that record's shape — the InsurTech sibling of the medical-adverse-event-incident-card (HealthTech), ai-student-record-incident-card-profile (EdTech), and title-chain-evidence-incident-card-profile (PropTech).

Event type taxonomy

Code Description
ai-unfair-discrimination-pattern Pattern disproportionately disadvantaging a protected class (CO SB 21-169 / NAIC §3 implicating)
ai-individual-adverse-action-error Single AI-recommended adverse action overturned, indicating model error
ai-ecdis-input-defect External Consumer Data and Information Source defect (NY DFS CL 7 implicating)
ai-fcra-dispute-pattern FCRA dispute pattern traced to AI tool reliance on contested consumer report
ai-actuarial-soundness-defect AI rating/pricing model output lacks documented actuarial soundness
ai-third-party-vendor-control-failure Third-party AI vendor missed contractual control
ai-governance-program-gap Internal AI governance program gap
ai-consumer-notice-defect AI-related consumer notice defect
ai-data-breach-with-ai-component Data breach with AI training-data or model-output component
ai-state-doi-exam-finding State DOI market-conduct exam finding implicating AI

Severity scale

Code Description
S1-low Single-consumer notice-only; remediated within 30 days; no regulator notification
S2-moderate Pattern < 100 consumers OR single material; state DOI notification recommended
S3-serious Pattern 100–1000 OR four-fifths violation OR ECDIS defect; state DOI notification REQUIRED
S4-critical Pattern > 1000 OR cross-state OR systemic four-fifths OR CO §10-3-1104.9 referral; multi-state DOI + CFPB + DOJ evaluation REQUIRED

Regulator referral pathways

state-doi-notification · naic-mcas-flag · cfpb-fcra-referral · doj-civil-rights-referral · state-ag-civil-rights-referral · state-doi-bulletin-violation-filing

Each incident records the evaluation status (filed / scheduled / evaluated-not-applicable / evaluated-not-required) per pathway, with destinations + timestamps + reasons.

Canonical example

examples/coastguard-aian-rural-incident.json — Coastguard Insurance's S3-serious unfair-discrimination pattern: VendorI ClaimsTriage v3.x 0.61 selection-rate ratio for AIAN-claimant rural property claims, 47 consumers affected across CT/NH/VT/RI, four state DOI notifications filed, NAIC MCAS scheduled, CFPB / DOJ / state-AG pathways evaluated and recorded as not-applicable / not-required with reasons. Cross-references the insurance-decision-record-audit-stream sibling (linked audit events) and the insurance-applicant-bias-coverage-lab sibling (linked bias coverage bundle).

Composes with

Repo Role
insurance-decision-record-audit-stream Source of linked_audit_stream_event_ids
insurance-applicant-bias-coverage-lab Source of linked_bias_coverage_bundle_id
naic-ai-bulletin-readiness-evidence-bundle Broader readiness bundle this incident becomes a part of
state-insurance-ai-disclosure-tracker Determines which state's bulletin-violation-filing pathway applies
medical-adverse-event-incident-card Sibling HealthTech Incident Card
ai-student-record-incident-card-profile Sibling EdTech Incident Card
title-chain-evidence-incident-card-profile Sibling PropTech Incident Card

Compliance posture

InsurTech-readiness scaffolding for the Incident Card record an insurer maintains for AI-implicating consumer harm events. Supports an insurer's program toward NAIC AI Model Bulletin §3 incident-response readiness, state DOI notification readiness, NAIC MCAS supplementary-disclosure readiness, CO SB 21-169 §10-3-1104.9 referral pathway readiness, and CFPB / DOJ / state-AG referral evaluation discipline. Does not by itself establish compliance with any state DOI bulletin or any federal regulator. Per the standing public-language guardrail: readiness · evidence · posture · controls · scaffolding — never "incident-response-compliant" or "regulator-attested" without an external attestation.

License

MIT — see LICENSE.

About

InsurTech AI Incident Card profile. 10 event types + 4-tier severity scale + 6 regulator-referral pathways with evaluation-state tracking. Covers CO SB 21-169, NAIC AI Model Bulletin governance gaps, NY DFS Circular Letter 7 ECDIS defects, FCRA dispute patterns, state DOI exam findings.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors