TL;DR

Bogus claims detected by Zurich UK reached a record £109.3 million over 12 months, Insurance Business reported on 5 November. AI features on both sides of the story, though unevenly: connected analytics found about a fifth of the cases, while AI-generated evidence is one tactic among several named by the insurer’s fraud chief. The figures show more detection, not necessarily more fraud, and they come from the company itself.

What Zurich reported

The total covers 4,359 claims, up 22% on 2022 in volume, or roughly one every two hours. Casualty made up the largest share at £61.6 million, followed by motor at £28.1 million and property at £19.6 million.

Using Quantexa’s analytics, Zurich checks each claim against sources including Equifax, Companies House and the industry’s fraud register. That accounted for 860 cases, worth £17.9 million. It says the approach has surfaced aliases, seemingly separate claims tied to one accident location, and networks that include rogue solicitors. The assertion that no other insurer connects claims this thoroughly is Zurich’s own.

How much is really AI?

The headline on the trade report points to gangs turning to AI, but the evidence given is narrower. Scott Clayton, who leads claims fraud work at Zurich UK and chairs the industry’s fraud bureau, lists AI-generated evidence alongside staged accidents, ghost broking and fabricated claims. Ursula Jallow of the bureau is reported as calling fraud increasingly AI-driven, without giving a figure. Neither puts a number on how many cases involved AI-made material.

That gap matters for anyone setting budgets. The Association of British Insurers recorded £1.34 billion of detected fraud across the sector in 2025, a rise of 14% in value even as cases dipped slightly. Our coverage of UK pharmacies and doctored photos shows a similar concern in another regulated sector.

Looking Forward

Insurance Business draws two lessons for brokers: warn customers about fake intermediaries, since ghost broking threatens reputations, and expect inaccurate disclosure to be caught more often as insurers join up their data. In our view, the second has a consumer edge. Linked data that speeds payouts for low-risk customers also widens what an insurer can see, and wrongly flagged customers will need a way to challenge it. Insurers publishing fraud records should report how many flagged claims turned out to be genuine, not only how many were stopped.