How AI-enabled fraud is reshaping global risk assessment in insurance

Generative AI has changed the rules of the road for insurers: photos, invoices, identity documents, and even voice recordings can now be fabricated or “improved” quickly and cheaply, from anywhere in the world, with fraudsters masking their location and operating across jurisdictions.

The result: not only is the volume of fraudulent applications increasing exponentially, ever-more advanced AIs are making the fakes hard to detect. This is forcing insurers across the globe to reassess how they validate evidence, and make checks and balances more robust, all without making the claims journey even harder for genuine customers.

In recent research from Verisk, a global insurance data analyst, 98% of insurers in the US market agreed that AI-powered editing tools are fuelling a rise in digital media fraud, and insurers reported much lower confidence in spotting deepfakes than spotting non-AI based edits to real photos or videos.

The fraud toolkit is evolving, not just expanding

In many cases, what we are seeing is not an entire loss being invented from scratch; it is “inflation” — taking a real event and making the damage look worse, or the supporting documentation look more credible, or the narrative provided with the claim appear tighter and more consistent than the underlying facts would otherwise support being the case.

The most obvious example is imagery. AI tools can generate convincing damage to a specific vehicle or property, producing photos that look realistic enough to pass a superficial review. Pair that with a repair estimate that appears to come from a genuine source and in an expected format — but has been overlaid with falsified data that can be created using easily accessible software tools — and you can create a claim file that looks complete and very convincing.

Invoices, receipts and estimates can also be manipulated using common editing tools, and some fraudsters are sophisticated enough to use layered editing in PDFs so that initial checks do not reveal what has been changed.

At the more advanced end of the scale, false identity documents are circulating, and new social engineering tactics, such as deepfake voice calls, are being deployed — for example, fraudsters can use synthetic speech modules to mimic local language, accents and dialogue. That makes it possible for a criminal based anywhere in the world to sound like they are from a certain region, adding an extra layer of believability to their cover story.

Why “enhanced evidence” is now a global risk-assessment issue

Fraudsters and AI have no boundaries. The same set of tools can be used from anywhere, and the same tactics can be trialled against insurers in multiple markets, at the same time, allowing the fraudster to simply hammer away until they find a chink in the armour — a protocol not followed, an assessor that doesn’t have time to properly give due diligence to each claim that comes across their desk.

That has direct implications for risk assessment, particularly for high-volume claims operations and any insurer relying on digital submission as the default route. Instead, insurers need to think in terms of provenance and integrity: where did the evidence originate, how consistent is it with broader data in that region, and does it stand up to specific manual scrutiny. Easier said than done for insurers processing hundreds, if not thousands, of claims a day.

But the industry can’t afford to just accept an increasing amount of AI fraud as ‘collateral damage’. Fraudsters are always looking for the path of least resistance — they want insurers to simply develop a higher tolerance to fraud as a cost of doing business. That cannot be allowed to happen.

The answer is to be even more disciplined: follow the right protocols, preserve evidence properly, and take advice early on the best route forward, whether that is civil recovery, repudiation, or referral for criminal enforcement.

Fighting AI with AI

While fraudsters are adapting quickly to the opportunities presented by AI, it is not true to say that insurers are stuck in the past. Many are investing heavily in tools that can screen for anomalies and signs of manipulation in real time, and that progress is essential. But I do not believe we can outsource judgement to technology entirely.

AI can help flag the cases that merit attention; it cannot, on its own, tell you that a claim is fraudulent. There is still no substitute for highly skilled intervention from well-trained individuals, when it comes to that final judgement call — is this real, or is it fake?

The mistake insurers must not make, is to think that beefing up AI solutions is a trade-off that means less investment in human training is needed.

Collaboration is the only sustainable route

The global, cross-jurisdiction nature of AI fraud, means that one insurer acting alone will always be playing catch-up. The tactics will move too quickly, and the fraudsters will simply shift to whichever organisation is least prepared.

That makes cross-border collaboration within the insurance community non-negotiable. In the UK, insurers already work through industry bodies and enforcement partnerships; internationally there are also investigator networks that share emerging trends and counterstrategies. It is only ever a matter of time before what is happening in one country, is happening everywhere so intelligence sharing is vitally important. The recent publication of the UK Government Fraud Strategy 2026 – 2029 reinforces the importance of cross sector industry collaboration and partnerships, intelligence sharing and fraud prevention.

That could require quite a culture shift in what is a highly competitive sector. Fraud prevention cannot be seen as a way of achieving competitive advantage — it has to be viewed as a baseline protection for the market and for honest policyholders.

The global nature of AI fraud also makes water-tight governance processes essential. Cross-jurisdiction intelligence sharing, while essential, can be incredibly complex.

Different jurisdictions have different requirements for privacy and information sharing, and it is entirely possible for a strong fraud case to unravel later if evidence handling and governance laws in one country haven’t been adhered to.

Priorities for insures and brokers

AI fraud isn’t going anywhere. While no insurer can ever be immune, there are practical steps they can take to reduce exposure.

The aim should still be a fast, fair claims process with targeted scrutiny where risk indicators justify it. Those justifications need to be supported by stronger evidence validation, better-trained people, and governance that stands up when decisions are challenged.

Recent legislation under the Economic Crime & Corporate Transparency Act 2023 and the impending Crime & Policing Act 2026, places additional corporate responsibilities on insurers; fraud prevention and risk mitigation must be a Board priority with an anti-fraud culture being driven from the top down.