Artificial intelligence is genuinely reshaping parts of how fraud works in insurance. It’s also not changing nearly as much as some of the coverage suggests. For fraud teams trying to make practical decisions about tools, workflows and priorities, the distinction matters.
The Australian data on AI and insurance fraud tells a more specific story.
The part that is shifting: volume, quality and speed
The most meaningful change AI has introduced into insurance fraud isn’t a new category of fraud. It’s a new scale of existing fraud, and a faster rate at which new methods spread.
Deepfake technology has made it considerably easier to produce convincing images, documents, and recordings. Swiss Re’s SONAR 2025 report flags that insurers are already seeing a rapid rise in deepfakes used in claims, with personal property lines particularly exposed (2). Motor, home contents, and travel claims sit at higher risk partly because they often involve a single claimant without independent witnesses, which means there’s less natural cross-referencing already built into the process.
The more subtle shift is in how quickly fraud tactics travel. ICA’s counter-fraud lead Andrew Gill put it plainly last year: “If your systems fail, you will be exploited quickly and, in the modern world, news of systems failing can spread incredibly quickly.” (1) That’s not a hypothetical. Online communities mean that a tactic that works in one market or one line of business can reach organised groups in another state within days.
What this changes practically is the pressure on validation. When evidence is easier to fabricate, the skill of verifying what’s in front of you, rather than simply assessing whether it looks plausible, becomes more important, not less. That’s a process and training question as much as a technology one.
The part that isn’t: why fraud happens, and what catches it
The motivations behind insurance fraud haven’t changed. Financial pressure, perceived opportunity and a calculation about the likelihood of getting caught are the same factors they’ve always been. AI lowers the barrier to certain types of fraud, particularly around document and image manipulation, but it doesn’t create new fraudulent intent where none existed.
The investigative instincts that catch fraud are also fundamentally unchanged. Experienced investigators notice inconsistencies: in timelines, in the way a story is told, in the relationship between documents and claimed events. AI tools can surface more data points, but they don’t replicate the judgment involved in knowing which inconsistencies matter and which don’t. That remains a human skill, and one that’s genuinely difficult to systematise.
The other thing that hasn’t changed is the basic importance of documentation, even though the pressure on it has increased. AFCA received 111,373 complaints in 2025, a 14% increase on the previous year and the highest annual volume in its history. Motor vehicle insurance generated 12,879 complaints, up 18%, with claim handling delays the single most complained-about issue across the entire financial system (3). Fraud teams don’t operate in isolation from that environment. A claim under investigation is also a claim that can generate a complaint, and the quality of the records behind a decision, what evidence was reviewed, what raised concern, and why a particular outcome was reached, is what makes that decision defensible when it’s later scrutinised.
The gap worth paying attention to
ICA members detected $560 million of opportunistic insurance fraud across motor and property in 2023. Undetected fraud is estimated to cost the industry around $400 million annually, a figure Gill himself has described as “quite a conservative estimate.” (1)
That gap between detected and undetected fraud isn’t primarily a technology gap. It’s an intelligence gap: the difference between what a system is built to flag and what a skilled investigator with the right information can actually uncover. Better AI tools may shift some fraud from one column to the other, but the size of that gap points to something structural. Workflows built around the fraud of several years ago, validation processes that haven’t kept pace with how evidence can now be manipulated, and investigative capacity stretched by complaint volumes and claim cycle pressure at the same time.
Closing that gap takes more than a new detection platform. It takes a clear view of where your current process has blind spots, and the willingness to address those specifically rather than assuming better tooling will resolve them.
kbs Intelligence works with insurers in Australia and New Zealand on fraud detection, investigation support, and workflow development. If you’d like to talk through what any of this means for your team, get in touch.
References
Australian Financial Complaints Authority — AFCA complaints hit record high in 2025
https://www.insurancebusinessmag.com/au/news/breaking-news/afca-complaints-hit-record-high-in-2025-claim-delays-was-the-no-1-issue-566622.aspx
Insurance Council of Australia — ICA probe to detail fraud menace facing industry
https://www.insurancenews.com.au/daily/ica-probe-to-detail-fraud-menace-facing-industry
Swiss Re SONAR 2025 — How deepfakes and disinformation amplify insurance fraud
https://www.swissre.com/institute/research/sonar/sonar2025/how-deepfakes-disinformation-ai-amplify-insurance-fraud.html
