How so? None of the fraud prevention APIs I've seen tell you if a given number is connected to a base station, only whether it's theoretically GSM, landline, VoIP, or something else. With the ability to tunnel LTE over an unstructured Internet connection ("wifi calling,") I don't have to be within range of a single US-based base station to "appear" like I am in the US to the mobile network.
And everyone tells me that only accepting "real" mobile phone numbers means they're just piggybacking on the identity checks that mobile carriers do. Which, if true, I'd like to introduce you to Constable George Crabtree and his nineteen perfectly valid mobile numbers, all of them with his name in the CNAM field.
All this does is prevent someone who's not tech savvy and who might be trying to save a bit of cash by using a Republic Wireless or TextNow or some other "free calls and texts!" service from fully participating in this new app-based reality we've constructed. Someone who actually wants to commit fraud will step right over these dumb speed humps and do whatever they like.
I'm not saying don't do fraud detection, I'm saying don't do fraud detection that is so screamingly trivial to bypass.