State DOI market-conduct exams touching AI are the 2026 enforcement frontier. NAIC AISET deploys nationally over 2026-2028; state-specific bulletins layer on top - New York DFS Circular Letter 2024-7, Colorado Reg 10-1-1, Connecticut MC-25-8, Nevada Bulletin 24-006, California CDI scrutiny under §790.03, Texas TDI evolving posture, Florida DFS aggressive examination practice. Each state has distinct examination protocols, request patterns, citation frameworks, and remediation expectations. NY DFS leads on proxy-test sophistication and aggressive enforcement timing. Colorado leads on inventory and bias-testing rigor with the July 1 annual compliance report. Connecticut and Nevada layer NAIC §4 onto state-specific consumer-protection focus. California combines DOI bias scrutiny with the Cumis/Brandt/§790.03 civil exposure. Texas and Florida emphasize claims-handling exam practice with developing AI guidance. Carriers facing market-conduct exams need state-specific playbooks: what each exam looks like operationally, what each examiner requests, what each state cites, and what the Colorado Reg 10-1-1 compliance report due July 1, 2026 actually contains. This lesson is the seven-state playbook plus the Reg 10-1-1 template, the sample examiner interrogatories, and the response patterns.

New York DFS Circular Letter 2024-7 and Exam Practice

NY DFS Circular Letter 2024-7 issued July 11, 2024 is the most aggressive state-DOI AI guidance through 2026. The Circular establishes: governance expectations for AI Systems used in underwriting and pricing; a "proxy test" requiring carriers to evaluate whether AI inputs serve as proxies for protected classes; affirmative disclosure expectations when AI is materially involved in consumer-affecting decisions; broker and producer obligations. NY DFS examinations leverage the Circular as the operational framework - examiners cite Circular section by section in information requests and findings.

NY DFS examination practice: examination opens with broad request for AI System inventory, §4-equivalent governance documentation, proxy-test analysis for in-scope systems. Examiners drill on specific Tier 1 systems with intensive proxy-test scrutiny. Sample examiner interrogatory: "Identify each AI System used in underwriting personal auto, homeowners, or specialty lines affecting New York policyholders. For each, provide the proxy-test analysis as defined in Circular Letter 2024-7 §III, including the variables tested, the methodology applied, the statistical thresholds used, and the conclusions reached. Where proxy concerns were identified, provide the remediation actions taken." Response pattern: carrier produces system list from registry, proxy-test memo with documented methodology, registry references for each system. Carriers without operationally complete §4 program face citation under Circular.

NY DFS enforcement posture combines consent orders with monetary penalties, public bulletins, and remediation requirements. NY DFS-issued consent orders become reference for other state DOIs; the New York posture shapes national enforcement direction. Brokers handling New York consumers face producer-license implications; broker E&O carriers price NY exposure separately.

Colorado DOI Reg 10-1-1 and the July 1 Compliance Report

Colorado DOI Reg 10-1-1 is the most structurally rigorous state-DOI AI regulation through 2026. Original Reg 10-1-1 covered life insurance use of External Consumer Data and Information Sources (ECDIS) with bias-testing and disclosure requirements. October 15, 2025 expansion covered private passenger auto and health benefit plans. First annual compliance report due July 1, 2026 - the operational deadline that defines Q2 2026 regulatory preparation for Colorado-licensed carriers.

Reg 10-1-1 compliance report template content: (1) algorithm inventory for in-scope lines (life, personal auto, health) with model name, accountable executive, risk tier, deployment scope; (2) bias-testing summary by model - methodology, results, remediation status; (3) ECDIS inventory for life insurance - external data sources feeding models with provenance, fairness testing per source; (4) consumer-facing disclosure status - what disclosure language is provided to consumers about AI involvement in decisions; (5) governance documentation - committee structure, accountable executives, attestation; (6) incident history - material adverse-impact incidents and remediation. Report submitted to Colorado DOI; published portions become public record under state law.

Sample Reg 10-1-1 examiner interrogatory: "For your personal auto AI pricing model deployed in Colorado, provide the most recent bias-test report (Quarterly per §4 governance), the disparate impact ratio with statistical confidence intervals, the BISG methodology and application, the equality-of-opportunity test results, the calibration parity analysis by protected class, and the remediation history including the Hispanic-surname cluster incident response if applicable to your portfolio." Response: carrier extracts from registry; produces test report; produces methodology documentation; references §4 program documentation. Carriers operationally complete on §4 produce response by extraction; carriers without produce gap-laden response.

Colorado examination practice: collaborative with carriers maintaining open communication; punitive with carriers exhibiting documentation gaps or pattern of incidents. Colorado DOI uses Reg 10-1-1 violations as referenced findings; remediation orders include enhanced testing cadence, third-party audit requirements, customer remediation obligations. Colorado DOI's posture has shaped Connecticut, Nevada, and 2026-2027 state bulletins.

Connecticut MC-25-8 and State-Specific Extensions

Connecticut Bulletin MC-25-8 layers NAIC §4 framework with state-specific consumer-protection focus. Connecticut Insurance Department examination practice emphasizes claim-handling AI scrutiny - Tractable estimates, CCC litigation prediction, Shift fraud referrals. Examiners request claim-decision documentation with AI involvement traceable to individual claims. Sample interrogatory: "Provide claim file documentation for [N] randomly selected denied claims from [date range] showing AI involvement in coverage determination, with file notes documenting human reviewer's verification of AI output, reason codes separating AI signal from human decision rationale, and consumer communication."

Connecticut response pattern: carrier identifies AI-involved claims through registry-claim linkage; pulls claim files with file notes; produces traceability documentation. Carriers without file-note discipline aligned to AI involvement face documentation gap. Connecticut DOI's posture on remediation: required claim re-review, consumer remediation on affected claims, enhanced training requirements.

Nevada Bulletin 24-006 and Emerging State Patterns

Nevada Bulletin 24-006 issued mid-2024 follows the NAIC §4 framework template with Nevada-specific adjuster and producer-license overlay. Nevada Division of Insurance examination practice emphasizes accelerated UW AI for life insurance and AI involvement in adjuster decision-making. Sample interrogatory pattern: "For your accelerated UW model deployed in Nevada, provide knockout reason codes, FCRA workflow integration documentation, ECDIS sources used, and consumer-facing adverse-action notice template. For your claims AI tools (Tractable, CCC, Shift, Hi Marley, Five Sigma) used in Nevada, provide adjuster-license affirmation that licensed adjusters retain personal judgment and AI is decision support not decision substitute."

Nevada response: carrier produces accelerated UW model registry entry, ECDIS inventory, adverse-action workflow; produces adjuster training documentation and file-note discipline. Nevada DOI's posture: aligned with NAIC but with adjuster-license focus; remediation orders may include enhanced adjuster training requirements.

California CDI and the §790.03 Civil Exposure

California Department of Insurance is the most active 2026 state DOI on AI bias scrutiny. CDI examinations combine DOI enforcement (consumer protection focus, bias testing, regulatory referral) with §790.03(h) civil-remedy exposure (unfair claim practices via AI bias). California producer license-board posture aligns with CDI scrutiny; broker E&O exposure compounds. Sample CDI interrogatory: "For each AI System affecting California policyholders, provide the bias-test methodology, results by protected class as defined under California law (including national origin, ancestry, religion, sex, gender, sexual orientation, age, disability), and remediation status. Identify any system producing disparate impact and document corrective action."

California response pattern: carrier produces fairness test results with California-specific protected-class coverage; documents remediation; addresses §790.03(h) civil-remedy posture in coordination with General Counsel. California protected-class scope is broader than federal - California examination explores categories federal law does not require carriers to test. Without California-specific testing, carrier faces gap that triggers DOI finding and §790.03 civil exposure.

California Cumis counsel implications: AI-involved coverage determinations may trigger Cumis where carrier-insured conflict exists. CDI examination practice may probe Cumis triggers; documentation of human review and disclosure to insured supports defense.

Texas TDI and Florida DFS Claim-Handling Emphasis

Texas Department of Insurance examination practice through 2026 emphasizes claim-handling AI scrutiny under Texas Insurance Code §541 bad-faith framework. TDI examiners review AI involvement in claim denials, reserve adequacy, fraud referrals. Sample interrogatory: "For denied claims from [date range], identify AI involvement, provide file notes documenting human reviewer verification, reason codes separating AI from human decision rationale. For Shift Technology fraud referrals, identify the false-positive rate by demographic class and document remediation if disparate impact identified." TDI 2026 advisory opinion on AI in claim handling is in development; carriers monitor TDI bulletin index for finalized guidance.

Florida DFS examination practice emphasizes Civil Remedy Notice (CRN) framework integration with AI scrutiny. Florida examiners coordinate CRN demands from plaintiffs with examination findings - patterns surfaced in litigation surface in examination. Florida 11th and 17th Circuit decisions in 2026 are starting to address AI specifically; examination practice incorporates trial-court precedent.

Both states' response patterns emphasize file-note discipline, claim-by-claim traceability, and §4 program documentation. Carriers without operationally complete §4 face citation and remediation orders.

Sample Examiner Interrogatories and Response Patterns

Cross-state examiner interrogatory patterns and response templates.

Interrogatory pattern 1 - Algorithm inventory request: "Provide a complete inventory of AI Systems in production affecting [State] policyholders, with risk classification, accountable executive, in-production date, last fairness test date, last drift test date, and vendor or internal-build status."

Response: registry extract with state filter. Format as table with one row per system. Narrative wrap explains methodology, definitions, scope decisions. Reference appendix for §4 program documentation supporting the inventory.

Interrogatory pattern 2 - Specific Tier 1 system deep-dive: "For the [Akur8 / Munich Re / Shift / Tractable] system, provide complete documentation including model card, training data lineage, fairness test results for last 12 months, drift monitoring data, vendor §4 attestation, SOC 2 Type II report, sub-processor disclosure, incident history, and version control log."

Response: extract registry entry contents; supplement with peer-review report, vendor scorecard, vendor SOC 2 report. Length per system: 8-15 pages with attachments.

Interrogatory pattern 3 - Bias-test methodology: "Describe the bias-testing methodology for [Akur8 personal auto pricing / Munich Re accelerated UW / Shift fraud detection]. Include statistical thresholds, demographic categories tested, BISG methodology and application, equality-of-opportunity tests, calibration parity, and remediation criteria."

Response: methodology document referencing applicable bulletins (Colorado SB 21-169 quantitative methods, NY DFS proxy test, ASOP 56 peer review). Sample results from most recent test cycle. Documentation of when criteria triggered remediation.

Interrogatory pattern 4 - Consumer remediation: "For any AI-related incident producing consumer impact in [State] during [date range], provide affected consumer count, remediation actions, customer communication, and current status. Include the [Hispanic-surname false-positive cluster / Akur8 drift event / Tractable estimate dispute pattern] if applicable to your portfolio."

Response: incident registry extract; remediation log; customer communication templates; current status. Reference §4 incident response runbook.

Examination Response Coordination and Timing

State DOI examinations operate on different timelines than AISET. Initial information request typically 30-60 days response; follow-up interrogatories often 10-20 days; document production may be 60-90 days for complex requests. Carriers coordinate response through CCO with cross-functional input similar to AISET coordination - Chief Actuary, Chief Claims Officer, CUO, CISO, CDO, GC.

Multi-state coordination challenge: carriers writing in NY, CO, CT, NV, CA, TX, FL face overlapping examination requests with different timelines, different formats, different emphasis. Coordinated extraction from §4 program and registry produces consistent content across state requests; state-specific framing handles state-specific emphasis (NY proxy test depth, CO Reg 10-1-1 inventory rigor, CA broader protected-class coverage, TX/FL claims focus).

Post-Examination Findings and Remediation Cycles

State DOI examination findings typically include: factual findings (what was observed), regulatory findings (what was cited), required remediation (corrective actions), monetary penalties (where applicable), public disclosure decisions (consent order, public bulletin, confidential treatment). Carriers respond to findings with: agreement and remediation plan, or disagreement with documented basis (rarely successful), or settlement discussions for material findings.

Remediation cycles: typically 90-180 days for implementation, 12-24 months for verification of sustained compliance. Carriers without operationally complete §4 face longer remediation cycles because remediation effectively builds the program; carriers with program face shorter cycles because remediation tightens existing operations.

Cross-state pattern: findings in one state often referenced by examiners in other states; remediation in one state demonstrates compliance posture relevant to other states' examinations. Strong multi-state coordination through CCO and General Counsel maintains consistency.

Key Takeaways

• Seven states with active 2026 AI examination practice: NY DFS, Colorado DOI, Connecticut DOI, Nevada DOI, California CDI, Texas TDI, Florida DFS. Each has distinct examination protocols, request patterns, citation frameworks, remediation expectations.
• NY DFS Circular Letter 2024-7 (July 11, 2024) is the most aggressive state-DOI AI guidance through 2026. Proxy test, affirmative disclosure, governance expectations. Examiners cite Circular section by section.
• Colorado Reg 10-1-1 compliance report due July 1, 2026 covers life, private passenger auto, health benefit plans. Algorithm inventory, bias-testing summary, ECDIS inventory for life, consumer disclosure status, governance documentation, incident history.
• Connecticut MC-25-8 layers §4 with state-specific consumer-protection focus. Claim-handling AI scrutiny - Tractable estimates, CCC litigation prediction, Shift fraud referrals. Claim-by-claim traceability required.
• Nevada Bulletin 24-006 emphasizes accelerated UW AI and adjuster-license discipline. Aligned with NAIC §4 with Nevada-specific adjuster overlay.
• California CDI combines DOI bias scrutiny with §790.03(h) civil-remedy exposure. Broader California protected-class coverage; Cumis counsel implications on AI-involved coverage determinations.
• Texas TDI and Florida DFS emphasize claim-handling AI scrutiny. Texas §541 bad-faith framework; Florida CRN coordination with examination findings; trial-court precedent integration.
• Examiner interrogatory patterns: algorithm inventory, Tier 1 deep-dive, bias-test methodology, consumer remediation. Response from §4 program and registry extraction; mechanical build passes; memo-only fails.
• Multi-state coordination through CCO and General Counsel maintains consistency across overlapping examinations. Findings in one state referenced in others; remediation in one state demonstrates compliance posture for others. Strong §4 program reduces remediation cycle length.