Insurance must prepare for a rise in deepfake AI fraud
01/22/2024
Scott Clayton, Zurich's head of fraud, shares how rapidly developing technology and digital manipulation threatens to put additional pressure on claims handlers, as deception becomes harder to spot.
Shallowfake technology involves manipulating existing documents and images using editing software. Deepfake technology, meanwhile, is a form of artificial intelligence (AI) that can be used to generate convincing hoax images, documents, sounds and videos.
Shallowfake claims
Clayton, right, explained that Zurich was seeing more and more claims that have been manipulated with shallowfake technology. "This is becoming one of the most emerging threats from a counter fraud point of view," he said.
"There's software that is easy for fraudsters to access, which can manipulate images.
"For example, we have seen devices that if you were to draw a picture of a building, it could provide that building with impact damage or a broken window."
Total loss
Clayton explained that, currently, it is most prevalent in motor, where fraudsters are photoshopping registration numbers onto total loss vehicles.
"We have seen an increase in people locating total loss vehicles on salvage agents' websites and then implanting a registration number onto that car," he said.
"There are then claims made for that vehicle and a claims handler assessing this claim would take that at face value, that it is that actual vehicle."
Artificial claims
However, as well as a surge in fraudsters using shallowfake technology, Clayton said claims that have been completely manipulated with deepfake technology are also on the rise.
"In support of fraudulent claims, we receive completely fabricated repair invoices, engineer reports and repair estimates, so this trend is evolving into a completely false paper-based claim."
Clayton added that, in the past, criminals who intended to scam insurers, such as creating a crash-for-cash claim, would need to physically have vehicles crash into each other, but now technology allows them to produce a completely fake claim at home.
"People can now create a fraudulent claim entirely from behind their computer and extract significant sums of money because these cars are total losses.
"There are also often high value cars, and that's an area that is hurting us right now, but we have put measures in place to educate and combat these types of scenarios."
This is a "worrying trend" for the whole industry, he said, as AI technology being used to create fake claims is only one component of how AI can be abused.
"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," he said.
"Will it get to a point where you doubt everything that's in front of you?
"As an industry, we must ensure we have capable technology that can identify these types of claims."
Putting measures in place
Clayton explained that Zurich had basic balance validation checks to support claims handlers and help ensure they do not take claims at face value.
Claims handlers must always check and validate documentation they receive, for example they should contact vendors named on invoices to confirm documents are from that business.
"We also run more detailed searches around the vehicles and individuals involved, as well as looking at how we can deploy software to identify these cases quicker for us, but this can be a time-consuming and expensive process," Clayton added.
However, it is not just claim handlers who are being challenged by digital manipulation. Clayton explained how software providers are affected by this trend too.
"Many companies that provide software protection are telling a story about how AI has evolved. So, three years ago, if you were to ask a generative software programme to generate a specific image, it could do it, but it wouldn't be very convincing," he said.
"But now these providers are reporting that these images are getting harder to detect and this type of technology has really evolved to the point where it's becoming more foolproof."
Ramping up
Clayton added that deepfake claims had only started to emerge within the past year, but it has been "ramping up".
"We don't know whether it is one particular gang that is targeting Zurich or it's a modus operandi on the criminal network, like a franchise model that we see with crash-for-cash scams."
"But what we do know is that these individuals are very aware of the insurance claims process because they operate in a sophisticated manner and cover their tracks, which could suggest more of an organised element."
Clayton said that while Zurich was not aware of the full extent of these scams, it was working in conjunction with Insurance Fraud Bureau to help develop its understanding and find counter fraud solutions.
"This is part of a wider operation but, with any criminal investigation, you need hard evidence and that's what we are missing at the moment." he said.
"Therefore, we want to ensure we continue to pay properly presented claims quickly but validate what's in front of us.
"Eventually, we will screen all documents or photographs, and let software determine which ones are giving cause for suspicion, but there will always be human intervention."
