Hacker News Reader

Comment by notdang | Hacker News Reader

They don't approve the loans automatically. They go through your photos, social networks,calls, etc. to make sure it's a real account.

View on HN

New use for generative AI?

Yet another argument for empowering the phone as a user agent. All data should be mocked unbeknownst to the application. Faker should be built in for all of these collection points.

This sounds like a sorta fun thing to try to pull off…

That can be easily faked. Seems susceptible to sybil attacks.

It's surprisingly difficult to fake years worth of conversations with family and friends...

The loanshark can say "show me the conversation with your boss. Show me the conversation with your mom.". They can cross reference things between the person's memory and the conversations. They can download all the conversations and put them in a database so the same conversation set can't be used to take out another loan.

They can also call your contacts to confirm details - ie. 'can you confirm Mary is your daughter? Does she have 2 cats and a dog?'

They can do that, but are they actually doing that?

They would do that if they needed because people were spoofing the system. Or they would do what the traditional banks are doing: avoid lending to people without a proven track multi-year record, physical location to monitor, and assets to hold as collateral for a downpayment.

However, I'm curious about one thing. Everyone who suggest spoofing the system for the sake of privacy, do you believe people who do that... will actually also repay the loan they were given? Or would they just use a fake name & commit fraud?

Easily is relative. Any defense can be beaten. The question is whether or not it's worth beating it. There's no point in spending a million to crack a safe with a thousand in it. Well, unless you just want to $%^& whoever owns the safe.

Very sure that you can spoof all of the above with 2-3 hours of effort.

I wonder how much effort it would be to automate the process of creating real-looking data (via AI or public-domain data) and then watch the money roll in (or at the very least DoS these scams enough that they can no longer tell real loan applications from all the noise).

Reply on news.ycombinator.com