EIOPA Opinion on AI governance and risk management
An opinion published in 2025 by the European insurance authority, specifying how the sector's existing rules apply to AI systems not classified as high-risk by the European regulation, through a risk-proportionate approach.
The European regulation on artificial intelligence classifies as high-risk systems, in insurance, those used for risk assessment and pricing in life and health insurance. Other uses, motor or home pricing, claims handling, fraud detection, fall essentially under existing sector law. The European Insurance and Occupational Pensions Authority published in 2025 an opinion specifying supervisors' expectations for these systems, based on obligations already set by the Solvency II Directive and the Insurance Distribution Directive. The approach is risk-based and proportionate: the intensity of measures depends on the system's impact on policyholders and on the company. Expectations cover governance and accountability, with defined roles, fairness and non-discrimination, data quality, documentation and traceability, explainability adapted to each audience, human oversight, and robustness and cybersecurity. For European insurers, the opinion closes an apparent loophole: a system that escapes the high-risk category does not thereby escape supervision, and national supervisors use it as a grid in their inspections.
Textbook case: a motor insurer uses a learning model to route claims to automatic handling or to an adjuster. The system is not high-risk under the European regulation, but during an inspection the national supervisor asks, relying on the opinion, for the data documentation, fairness tests and human review procedure for routed files.
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.