Fraudsters turn to artificial intelligence

AI-enabled synthetic claims fraud is reshaping how insurers confront deceit, with the speed of digital manipulation now outpacing traditional detection methods.

Scale of the problem

In 2024 UK insurers identified roughly £1.16 billion in fraudulent general‑insurance claims, according to ABI research. Those false claims stemmed from 98,400 individual cases, a 12 % rise over the previous year. The fraud spanned intentional vehicle crashes, forged documents, ghost broking and, most commonly, modest exaggerations.

The sheer variety of tactics makes it difficult for insurers to pinpoint the exact frequency of synthetic fraud, a subcategory that relies on artificially crafted media and identities.

How AI fuels synthetic fraud

Matt Gilham, director at WhiteElk Fraud Performance Consulting, notes that fabricated evidence is not new, but AI has dramatically lowered the barrier to creating convincing falsifications. “What has changed is the speed and ease with which fraudsters can find, create or amend digital evidence,” he said, adding that the true prevalence of fraudulent documents remains uncertain.

Jon Bethell, head of private clients at Verlingue, describes AI‑generated fraud as the market’s most underestimated threat. He illustrated a scenario where an image of a living room is altered by AI to display water damage and a cracked roof, turning a routine claim into a fabricated disaster.

Even minor deceit can be amplified by AI. Bethell warned that a fake receipt generated by an AI tool could be inserted into thousands of claims each week. “If you fire a fake AI receipt into every single one of those claims, [that fake evidence] will get through because they’re really, really busy,” he explained, highlighting the pressure on claim teams handling high volumes.

Verifying every low‑value claim with an on‑site inspector is not economically viable. While loss adjusters will examine high‑value submissions, many claims around £2,000 are processed without physical verification, creating a gap that AI‑crafted evidence can exploit.

Industry response and lingering obstacles

Insurers are bolstering image‑verification capabilities and exploring document and video validation tools. Gilham says that metadata loss during media submission can hamper these solutions, leading to false positives or missed fraud. “False positives and false negatives still need careful management,” he noted, highlighting operational strain on teams already juggling multiple fraud indicators.

Another emerging challenge is claimants disputing insurer‑collected digital evidence, alleging it was AI‑generated or altered. This pushes insurers to strengthen the chain of custody for digital media, ensuring that evidence remains admissible and trustworthy.

On the defensive side, AI models can be trained to spot duplicated submissions, inconsistencies in metadata, and text that bears the hallmarks of machine generation. Such tools offer a way to counteract the very technology that fuels synthetic fraud.

In the broader sense, the situation mirrors past waves of fraud where new tools—once the domain of criminals—eventually become part of the defensive arsenal.

When electronic records first entered the insurance workflow, they sparked similar concerns about manipulation, yet over time verification methods adapted. The current AI surge follows that pattern, suggesting that while the threat is real, the industry may develop countermeasures as quickly as fraudsters innovate.

AI fraud grows fast.

Ultimately, insurers must stay vigilant. The balance between leveraging AI for detection and preventing its abuse will determine how effectively the sector can safeguard against synthetic claims fraud moving forward.