Regulation — New York State
NY DFS Insurance Circular Letter No. 7 (2024) — AI + external consumer data
New York Department of Financial Services issued Circular Letter No. 7 on July 11, 2024. It applies to every NY-authorized insurer using AI systems or external consumer data / information sources (ECDIS) in underwriting or pricing. NY DFS uses circular letters to state supervisory expectations that carry practical weight in market-conduct exams.
What the regulation actually says
"Insurers should develop and implement a written program for the responsible use of ECDIS and AIS … that includes governance, risk management, and internal controls, and includes senior management and board oversight."
NY DFS Circular Letter No. 7, Section III (July 11, 2024) — governance requirement
"Insurers should conduct testing to assess the potential for unfair or unlawful discrimination … before deploying an AIS or ECDIS in an insurance decision, and periodically thereafter."
NY DFS Circular Letter No. 7, Section III — unfair-discrimination testing
"The Department expects insurers to be able to explain to consumers … the specific reasons for an adverse underwriting decision resulting from the use of ECDIS or AIS."
NY DFS Circular Letter No. 7, Section IV — adverse-decision transparency
What this means in plain English
Three New York-specific obligations layered on top of the NAIC baseline:
- Written AI + ECDIS program. Same shape as the NAIC Bulletin, but NY DFS will audit against it during market-conduct exams.
- Pre-deployment + periodic testing. Testing timing is spelled out: before the AI or ECDIS goes live in NY, and on a recurring cadence after.
- Adverse-decision reason-giving. NY consumers can be told the specific ECDIS or AI factor that drove the decision. That maps to a decision-record retention duty on the insurer.
What triggers the exposure in the sample
I3 group decline / refer-SIU lane rate rose 11% to 62% silently. If any affected quotes originated in New York, three separate expectations under Circular Letter No. 7 are now in play: governance (was the AI in the written program?), testing (was pre-deployment + periodic testing done?), and reason-giving (can the specific ECDIS driver be produced to each affected consumer on request?).
What the $499 Snapshot shows against this rule
- Independent measurement of the group-differential pattern — the "periodic testing" evidence for the NY subset
- Model-version pinning + decision hash + input snapshot — the reason-giving reconstruction record
- Signed independent-verifier declaration — distinct model family + distinct retention pipeline consistent with Circular Letter Section III governance
- 3 fix-first items with named criteria — the remediation path NY DFS expects to see in the operator's file
See the 5-decision reproducibility drill →
$499 Snapshot. 3 business days.
Independent-verifier determination scoped to your NY-authorized AI surface + 3 fix-first items + signed declaration.
Buy $499
Snapshot credit applies to Baseline ($2,500) or Enterprise Attestation ($35-55K) upgrade within 30 days.