Can AI underwriting models use data sources beyond traditional risk factors

Expanding Beyond Traditional Actuarial Categories

Yes, AI underwriting models genuinely can and do incorporate data sources beyond traditional risk factors, reflecting a broader shift in the insurance industry toward using richer, more varied data to assess risk, though the specific data sources permitted and actually used vary considerably by insurance line and jurisdiction.

Examples of Alternative Data Sources in Use

Telematics data from connected vehicles — tracking driving behaviors like braking patterns, speed, and time of day driven — has become a widely used alternative data source in auto insurance underwriting and pricing. In some life and health insurance contexts, insurers have explored using wearable device data tracking activity levels and other health-related metrics, and some underwriting models have incorporated broader digital data sources as well.

Why Insurers Are Interested in This Kind of Data

These alternative data sources can potentially provide more direct, real-time insight into actual risk-relevant behavior than traditional demographic or historical categories alone, potentially enabling more accurate and individualized risk assessment — a driver, for example, might get more favorable pricing based on genuinely safe driving behavior directly observed through telematics data, rather than being priced based only on broader demographic categories.

Why Regulatory Requirements Constrain This Use

Despite this technical possibility, regulators generally require that any data source used in underwriting have a demonstrated, legitimate connection to actual risk, and some alternative data sources have drawn specific regulatory scrutiny over concerns that they might function as a proxy for protected characteristics, indirectly producing discriminatory outcomes even without directly using a protected characteristic itself.

Why Consent Requirements Often Apply

For data sources that require active customer participation, such as installing a telematics device or sharing wearable device data, insurers generally need to obtain informed customer consent before collecting and using this data for underwriting purposes, reflecting both regulatory expectations and practical necessity given the voluntary nature of this kind of data sharing.

Why Permitted Data Sources Vary by Jurisdiction and Insurance Line

Given differing state-level insurance regulation and differing rules across different insurance lines (auto, life, health, property), the specific alternative data sources an insurer can use, and how they can be used, varies considerably, meaning a data source permitted in one jurisdiction or insurance line might face restrictions in another.

Bottom Line

AI underwriting models genuinely can incorporate alternative data sources beyond traditional risk factors, including vehicle telematics and wearable device data, but the specific data sources permitted vary by insurance line and jurisdiction, since regulators generally require any data used to have a demonstrated, legitimate connection to actual risk and not function as a proxy for illegal discrimination.

Go deeper

• How Do Insurers Use AI to Assess Risk and Set Premiums?
• What Is Model Risk and Why Do Regulators Worry About AI Models in Banking?
• How accurate are AI crop yield predictions compared to traditional methods?