AI-Generated Insurance Fraud Is Getting Better, And No Single Test Will Catch It
Reported by Forbes1 source ↓
Insurers have entered an arms race, and the first step is admitting it. The second is accepting that no single tool will win it for them.
Key points
- The proposed approach combines image and document checks with policy histories, claims archives and external records
- The article says cleared claims should be labeled and fed back to models to reduce repeated false flags
- Aviva reported detecting a record £233 million in suspected claims fraud in 2025
- The NAIC AI bulletin had been adopted in 25 jurisdictions as of April 2026, according to the article
Details
He argues that insurers should combine image findings with other indicators, such as mismatched dates, recently registered repair shops, policy changes and similar claims in their archives. He cautions that individual anomalies do not prove fraud.
Nowak says investigators’ findings, including when a flagged claim is legitimate, should be returned to the system so models and thresholds can be adjusted. He also recommends regular audits of flagged claims, missed fraud and false positives, and says final decisions on suspicious claims should remain with people. The article cites the December 2023 NAIC model bulletin on insurers’ use of AI, saying it had been adopted in 25 jurisdictions by April 2026.
Written from coverage as of
Images
- Forbes