Insurance AI targets staged claims while document fraud drives most losses

UK insurers detected £1.16 billion in fraudulent claims in 2024, the latest year in a decade-long run above the billion-pound mark. The largest single category was not staged crashes or organised rings-it was exaggerated loss at £466 million: padded household lists, inflated repair invoices, and refactored medical bills. For every pound of fraud the industry catches, an estimated two pounds slip past undetected, according to the Association of British Insurers.

That gap is widest in soft fraud, which Deloitte's 2025 analysis puts at 60% of incidents with detection rates between 20% and 40%. Hard fraud-staged and premeditated-makes up the remaining 40%. EY's claims practice estimates total leakage from fraud, error, and scope creep at 7% to 14% of carrier claims spend.

AI investment is misaligned with the real leakage

From research of thirteen publicly named carrier AI deployments for claims-fraud detection in the past three years, only four lead with document-content analysis: Aviva's special investigations unit, Zurich Switzerland's German-language pilot, Swiss Re Corporate Solutions' ClaimsGenAI, and NFU Mutual via Synectics' SynDOC layer. The remaining nine are image-based, network-based, or voice-based, reflecting a broader industry trend in AI for Insurance.

Aviva's Peter Ward describes "a steady rise in manipulated images and documents supporting opportunistic claims." Zurich UK's Scott Clayton said, "I never thought it would happen in my career, that a document or an invoice or a photograph in front of you is entirely fictitious."

Generative AI has changed the fraudster's toolkit

Allianz UK reported that cases where apps were used to distort real-life images, videos, and documents increased by 300% between 2022 and 2023. Admiral's detected fraud rose 71% in 2025, partly attributed to AI-generated supporting evidence. BDO's May 2026 FraudTrack report notes fraud is "becoming more technologically enabled, more adaptive and more resilient."

The CVPR 2023 DocTamper benchmark produces F1 scores near 0.91 in-domain, but the same class of detector drops to F1 around 0.04 on a different receipt dataset out-of-distribution. A 2026 study of diffusion-model edits to financial documents, AIForge-Doc, found the best zero-shot detector scored AUC 0.75 against modern generative tampering. A model trained on Chinese receipts will not generalise to a Munich body-shop invoice. No peer-reviewed benchmark exists for document-fraud detection on European insurance documents in the languages European carriers actually run on.

Regulators are closing the explainability gap

The EU AI Act carves fraud-detection systems out of its high-risk regime, but other pressures are already enforceable. FINMA Guidance 08/2024 in Switzerland names insurance claims determinations as the highest tier of explainability. The EIOPA Opinion of August 2025 covers claims management and fraud detection across the European insurance value chain, while the NAIC Model Bulletin has been adopted in twenty-four US states. GDPR Article 22 gives every European claimant the right to demand human review of a solely automated decision. The March 2026 Lokken v UnitedHealth discovery order is the canary: courts will compel algorithm documentation and "the vendor said it was accurate" does not survive subpoena.

Reuben John is CEO and co-founder of True Aim, a Swiss insurtech building auditable AI for claims decisioning.

Why this matters for insurance professionals

Claims and fraud leaders should stop accepting in-distribution accuracy figures as evidence of production capability. Demand cross-domain evaluation in the languages of your book and performance against AI-doctored documents. Build detection systems that flag, cite the rule, name the human in the loop, and produce a record a regulator can read-the same discipline that catches paperwork-level fraud is what regulators are already demanding.