Claims AI. Underwriter evidence.

Insurance AI is being scrutinized by state regulators, class-action counsel, and model-risk examiners on three continents — often about the same decisions.

Where insurers need evidence today

AI in insurance is now regulated by a patchwork that would not have existed three years ago. The NAIC Model Bulletin on AI frames the direction for US state DOIs. Colorado's life insurance regulation requires testing for unfairly discriminatory outcomes. New York DFS Circular Letter 7 governs external data use in underwriting. The EU AI Act classifies insurance pricing and claims handling in the high-risk annex. Carriers operating across jurisdictions face overlapping obligations, each asking a version of the same question: how did this system reach this decision, and can you prove it?

The five use cases we focus on

Claims automation

AI-assisted claims decisions — fast-track approvals, fraud flags, coverage interpretation — are disputed in the same channels as traditional claims, but with the added question of whether the AI applied policy consistently across similar claims. Veridra produces a signed record per claim decision that an adjuster review, regulator audit, or class-action discovery can verify independently.

Pricing and underwriting

Rating decisions signed per quote, with the specific factors, scores, and external data sources hashed into the record. When a disparate-impact challenge arises, the signed decision stream allows exact reconstruction of how the rate was set — including which external data sources were queried, under which policy version, at which confidence level.

Fraud detection

Fraud-flagging decisions are a specific risk area: a false positive denies legitimate coverage, a false negative costs real money, and the decisions often happen at the edge without human review. Veridra signs each flag decision with the policy version in effect, the triggering pattern, and the disposition (investigate, deny, refer). Post-event audit reconstructs the decision exactly.

External data governance

New York DFS and multiple other regulators are focused on the use of consumer credit, health indicators, and lifestyle data in underwriting. Veridra records which external data sources were queried for each decision, which versions were in effect, and what disposition the model made of each input. Regulators asking what data did you rely on? get a cryptographically attested answer per decision.

Adverse action and appeals

When a customer disputes a decision, the adverse action workflow needs the original signed record, the policy version at the time, and the reasoning chain. Veridra produces this package in seconds rather than weeks. For appeals that reach regulators, the same package is admissible evidence.

Unique obligations for insurance

• Unfair Trade Practices Acts — state-level. Requires demonstrable fairness in AI-driven decisions. Watch module's continuous drift detection is a direct fit.
• Colorado Regulation 10-1-1 — life insurance AI cannot produce unfairly discriminatory outcomes by race, color, national origin, or religion. Requires testing and evidence.
• NAIC Own Risk and Solvency Assessment (ORSA) — AI risk is now material risk; governance and evidence are required annually.
• EU Insurance Distribution Directive (IDD) Article 20 — pricing transparency to customers, especially for automated or AI-assisted decisions.
• State-level rate filing requirements — many states require disclosure of AI use in rate-making decisions.