EU AI Act for Insurance: High-Risk Rules for Risk & Pricing AI
Insurers increasingly use AI to price policies and assess applicant risk — and the EU AI Act (Regulation (EU) 2024/1689) treats the most sensitive of these as high-risk. AI used for risk assessment and pricing in life and health insurance is expressly listed in Annex III. As with the rest of the Act, the rules follow the output into the EU market, wherever the insurer or its AI vendor is based.
Is it in scope?
Annex III, under access to essential private services, specifically covers AI systems intended to be used for risk assessment and pricing in relation to natural persons in the case of life and health insurance. That is a deliberately narrow trigger: it targets life and health lines, where an unfair model can deny people essential cover. Other insurance uses — such as claims-triage automation or general fraud detection — are not caught by this specific listing, though insurers should assess each system on its facts rather than assume a blanket exemption.
Typical AI use cases
- Automated risk assessment for life insurance
- AI-driven premium pricing for health cover
- Underwriting eligibility and acceptance decisions
- Applicant risk segmentation and scoring
- Claims triage and automated assessment
- Fraudulent-claim detection
Risk classification
AI for risk assessment and pricing in life and health insurance is high-risk under Annex III. An insurer that develops the model in-house is the provider and must meet Art. 9–15, with particular attention to data governance and bias (Art. 10) given the discrimination risk in pricing. Where the model is supplied by an insurtech vendor, the insurer is a deployer under Art. 26. Claims and fraud tooling generally sits outside this listing, but should be documented so the boundary between high-risk pricing and other functions is defensible.
Obligations to prepare for
FAQ
Which insurance AI is actually high-risk?
Annex III specifically covers risk assessment and pricing for natural persons in life and health insurance. Those systems are high-risk; the listing does not automatically extend to every other insurance use case.
Is motor or property pricing AI caught by this rule?
Not by the Annex III insurance listing, which is limited to life and health insurance. Other lines should still be assessed individually, but they don't fall under this specific high-risk trigger.
Why is bias such a focus for insurance AI?
Because pricing and risk models can indirectly discriminate against protected groups. Art. 10 data-governance duties require you to examine training data for bias and mitigate it, which is central to high-risk insurance compliance.
Run the free 2-minute risk checker — no sign-up.