I work in insurance servicing and I have been thinking about this recently. AI photo and video generation is getting scary realistic. How long before we see fabricated damage photos, fake dashcam footage, or manipulated documentation in claims? Are SIU teams prepared for this? Also what about legitimate claimants whose real evidence gets questioned because "it could be AI"? Curious if anyone in claims or fraud investigation is seeing this come up yet?
AI-generated evidence in claims, is anyone seeing this yet?
byu/RedBloodedGod inInsurance
Posted by RedBloodedGod
1 Comment
Somebody else posted on this topic a couple weeks ago with some interesting examples.
I periodically have AI generate me a picture of a certain vehicle with specific damage, and yes, it’s not good enough for fraud, but it’s getting pretty close. It seems to still miss a lot of details and seems to avoid trademark infringement, so stuff like emblems don’t match-up.
The photo ap programs get access to GPS and try to correlate the pictures with the expected time and location they’d be taken, and flag them for assorted fraud parameters.
There can be staff or services arranged to visit specific or random samples to verify things like physical damages, but I think the carriers will amortize the expenses in some cases, in lieu of hiring actual IRL field people to go look at stuff.
I definitely feel like I pay wayyyy more soft fraud since APD claims went heavily to virtual, and carriers have definitely lost the plot on labor rates when they gave up on field staff. The ships are running amok.
I’m actually expecting large MSO’s to start buying P&C insurers the way CVS bought Aetna.