Macquarie University cyber security experts have invented a multi-lingual chatbot designed to keep scammers on long fake calls to waste their time and ultimately reduce the huge number of people who lose money to global criminals every day.
Begun, the bot wars have…
Teaching AI to deceive? What could go wrong?
Named Apate, after the Greek goddess of deception, the system will 'scam the scammers," using convincing voice clones to conduct conversations with real scammers.
I half expected them to just call it Kitbogabot.
I really want to hear recordings of these conversations. I think it’d be fascinating.
https://aff.419eater.com/viewforum.php?f=29&sid=ba36f1c70b17dbbfaa5cc6315689443e
people have been scam baiting for ages. my favorite is the one where they got them to do the dead parrot sketch
Not the same thing, but an older site/service where they would just randomize various automated responses during gaps in the conversation to keep the person on the phone. They have a bunch of recordings.
Fark (possibly Something Awful, it was forever ago). iirc, had a run of scamming the Nigerian Prince email farms. It was pretty savage, having humans send humiliating/degrading time stamped/dated photos via email, which in turn ended up on Fark. The first couple of stories seemed funny in a disturbing kind of way, and about the third or fourth column, I was really uncomfortable. Then I read about how these farms worked by trafficked human power.
There’s also https://www.419eater.com/
I’m sure the scammers are working on chat bots that will do the scamming without the costly need for people.
Scammers have long since used bots already for text based scams (though dumb ones). Phone calls are a lot harder, though. And there’s also “pig butchering” scams, which are the long cons. Most commonly those are fake relationships. I suspect those long cons would have a hard time convincing someone for months, as human scammers manage to do.
I suspect that scammers will have a harder time utilizing AI, though. For one thing, scammers are often not that technologically advanced. They can put together some basic scripts, but utilizing AI is difficult. They could use established AI, but it’d almost surely be against their ToS, so the AI will likely try to filter scam attempts out.
That said, it might be just a matter of time. Today, developing your own AI has a barrier to entry, but in the future, it is likely to get a lot easier. And with enough advancements, we could see AI being so good that fooling someone for months may be possible. Especially once AI gets good at generating video (long con scams usually do have scammers video chat their victims).
And honestly, most scams have a hundred red flags anyway. As long as the AI doesn’t outright say something like “as a large language model…”, you could probably convince a non zero number of victims (and maybe even if the AI fucks up like that – I mean, somehow people get convinced the IRS takes app store gift cards, so clearly you don’t have to be that convincing).