AI scam callers can be broken with a few simple commands

AI is coming for all our jobs, it seems, even those of phone scammers. What is the world coming to when we can't even get swindled by good, honest (well, not really) humans anymore?

The advent of AI has massively improved the scalability of these kinds of scams, to the point where a human doesn't even need to be involved in the process. The upside? These models are still fragile enough to be broken with a few simple commands, sending them spiraling into endless loops and costing scammers precious time and/or money. Scambaiter Kitboga demonstrates this technique in flawless detail, sending one AI phone scammer after another to purgatory (or to Albuquerque, which is basically the same thing).

Try it yourself next time you get a spam call! Get them to burn through their tokens for the good of unsuspecting victims!

Previously: