Welcome To The Brave New World Of Vehicle Insurance Fraud Powered By Shallowfakes
from the I-can’t-believe-my-eyes dept
The harms of deepfakes have been evident for a while. Recent examples include a Biden deepfake designed to influence voters, and the rising use of AI “nudification” apps to produce deepfake nudes of students, often female minors. But alongside the application of sophisticated AI programs to produce deepfakes there are other, lower-tech scams, often known as “shallowfakes”. An article in the Guardian explains “shallowfakes can be created using conventional editing software on a phone and apps such as Photoshop.” This is leading to a “surge in fraud cases” according to the article, involving vehicle insurance claims:
[The insurer] Allianz, which includes the general insurance arm of LV=, said in one case an individual had a photo of his van posted on his social media page as part of his business and ended up having a claim pursued in his name for an accident that never took place.
LV= received images of his vehicle from the fraudsters that seemed to show the front bumper had been cracked in the alleged accident, along with a fake repair invoice for more than £1,000 [around $1,250].
The LV= fraud team investigated and found the photo was identical to the one on the social media page, except for the fact that the image had been doctored to show the fake damage.
The article mentions another example of how shallowfakes are being used. Fraudsters are finding vehicles regarded as total losses by insurers on sites selling them for salvage, and then placing a different license plate on the image using digital editing tools. False insurance claims are then made with the shallowfake vehicle image.
These low-tech images probably aren’t hard to spot, but moving from shallowfakes to deepfakes could make that more difficult. In any case, the rise of both kinds of manipulation underlines once more that we live increasingly in a world where images and videos can no longer be taken at face value.
Follow me @glynmoody on Mastodon and on Bluesky.
Filed Under: biden, cars, deepfakes, fraud, insurance, nudification, photoshop, shallowfakes
Companies: allianz
Comments on “Welcome To The Brave New World Of Vehicle Insurance Fraud Powered By Shallowfakes”
The big difference that AI brings to the table here is that with shallowfakes, a single person can spend 30 minutes creating a single fake image they can file for a claim.
With AI, the fraudster can generate 500 images, none of which are matches for anything available via a Google Image Search, but all can be keyed off the same vehicle, from different perspectives. They can also generate legitimate looking invoices, websites and even emails and form details that are all consistent with a single claim.
And all of this can be done with less than 5 minutes of the fraudster’s time (a bit more of the computer’s time, of course).
Re:
And of course, some of those deepfakes show vehicles with 5 wheels, or the grill of a 1965 T-bird (and the body of a 2021 Elantra).
Or, perhaps that the ’65 Skylark left tire marks showing that it had positraction.
They banned TikTok for less than that. Why they’re waiting to ban Photoshop?
As with a lot of things, the solution seems to be a traditional, low-tech one: return to insurance adjusters inspecting the vehicle as part of claims processing. The practice of approving claims based only on photos from the owner and/or shop inherently accepts a certain level of fraud, because humans, and if that’s becoming unacceptable then change the practice back to what it was.
Re:
Yeah…a couple of ruffles and flourishes on this idea: 1) onsite inspection would only be needed on a percentage of claims to discourage both the shallow and deep fakes. That would make the cost more reasonable. This would seem to be a good candidate for a legit AI application. 2) Prosecute the worst offenders with a lot of fanfare to magnify the impact of the investment.
This comment has been flagged by the community. Click here to show it.
"photoshop can be used for fraud"
Lol, the state of this site.
Not that that the more main stream “News” sites wouldn’t run the same lede, but they’d spend more owrd pretending it was interesting.
Re:
Well, the article is quoting The Guardian, if that’s what you’re referring to 🙄
Re:
Why bother with photoshop when a sharpie will do?
I know in my state, not sure about others, if there is a claim made, there needs to be a police report. If no police report there needs to be an inspection by an Insurance adjuster of the company’s choice
So there's no physical verification?
So, let me get this straight: there’s no more physical verification of the damage? Then of course people are going to take advantage of the system and sent in altered photos.
However, is this really something to get worked up about? I’m sure the insurance companies realized that it’s still cheaper to pay put fraudulent claims than pay employees to physically inspect the damage.
Re:
The average insurance payout for a totaled car is $6,250. Multiply that by the number of false claimants and I’m sure it’s likely to dwarf the average annual salary of $46,900 earned by loss adjusters. Better to physically verify the state of the vehicle, and if the claim is false, the company gets to stop insuring the vehicle and keep all premiums paid to that point. Much cheaper for the insurance company all round. TL;DR: DYFR.
We’ll just fire all of our adjusters who physically go an see the cars & just let people send us pictures… what could go wrong??