The Insurance of Things Versus the Insurance of People: Why AI Will Split Europe’s Two Insurance Worlds

By Paul Donnelly, Global Head of Insurance at Version 1.

Consider a woman in Lyon who survived cancer four years ago. She is healthy now, back at work, ready to buy her first apartment. Under the Loi Lemoine, the insurer providing her mortgage protection cannot ask a single medical question: the property is an ordinary home, the loan will be repaid within her working life, and that is that. Her brush with illness, the very data point that would once have doubled or tripled her premium, is invisible to the underwriter. It is a small, quiet victory for the individual, and it captures something Paul Donnelly believes the rest of the industry has been slow to grasp: in European life and health insurance, the data that matters most is increasingly data that insurers are no longer permitted to see.

This matters now because a fundamental divergence is opening between property and casualty (P&C) insurance on one side and life and health (L&H) on the other. Donnelly frames it simply: “P&C is the insurance of things, and L&H is the insurance of people.” In P&C, the data available to insurers is growing explosively. Sensors in buildings, vehicles and infrastructure produce vast quantities of information about physical risk, with few regulatory barriers to using it. A building fitted with leak detectors and environmental monitors can be profiled with extraordinary granularity. The trajectory is towards ever more data, ever finer pricing.

Life and health insurance is heading in the opposite direction, at least in Europe. The data involved is not about bridges but about bodies: medical histories, family health, lifestyle. When someone applies for life cover, they might mention a sibling’s brush with cancer. That information sits on the insurer’s systems for decades, and it is, as Donnelly points out, not even the applicant’s data; it belongs to the sibling. European regulation treats this seriously. 

GDPR established the framework, but more specific measures keep tightening the constraints. France’s right to be forgotten allows cancer survivors to omit their history from applications after a defined period, a provision now spreading across Europe with the qualifying window shrinking towards three years. The Loi Lemoine bars insurers from asking medical questions of ordinary mortgage applicants altogether. The EU Gender Directive restricts the use of sex as a rating factor, despite actuarial evidence that women live longer than men.

Each measure is, individually, a small adjustment. Taken together, they describe a world in which the data life insurers may use is contracting even as their technical capacity to analyse it expands. “In P&C, there’s an ever, ever-growing amount of data that insurers can use,” Donnelly says. “And on the other side, it is under duress.” The result is a drift towards mutualisation, pooled pricing in which the cost of risk is spread across populations rather than allocated to individuals.

Donnelly sees this as part of a deeper philosophical split. He identifies three global postures towards data: in China, state stability; in the United States, innovation and commercial freedom; in Europe, the protection of individual rights. That principle now shapes what AI can do in European life insurance in ways that consultants from retail or technology do not always grasp. “It is easy to walk in from another industry,” he warns, “and to kind of step onto these landmines.”

Back in Lyon, the woman signs her mortgage papers without ever having disclosed her illness. For her, it is simply fairness. For the insurer, it is one more risk absorbed into the pool, priced not by what is known about her but by what the law has decided should remain unknown. Donnelly does not begrudge the principle; he simply notes its consequences. The insurance of things is becoming ever more precise. The insurance of people, in Europe at least, could be moving the other way.

“In European life and health insurance, the data that matters most is increasingly data that insurers are no longer permitted to see.”