Fairly certain that AI (meaning an expensive llm type model) isn't needed to detect spam a large amount of the time. Classical classification methods could work while also being more privacy friendly (e.g. running on device).
> According to AT&T, a big difference with its product is that it is built into the network itself.
Counter-proposal: AT&T knows who is paying for the spam calls, even when they forge a caller ID. Show THAT COMPANY'S number and let customers sort it out...
FWIW they've gone through many iterations of this. In the 1960's there were rooms of operators listening for keywords. That was replaced with computers that listened for keywords and phrases which I think improved privacy. Whistleblowers appear every couple decades and disclose NSA listening locations. I am mostly curious if this AI at least understands context to reduce false alerts thus minimizing alert fatigue, wasted resources, etc... Bonus if it can troll some spammers. Bigger bonus if one day it can pit them against one another using psychological warfare.