Though, to be clear, "security screening" is a bit of a misnomer here. It isn't a one-time thing. The point of giving someone a security screening isn't really the pre-screening that occurs when you first sign up; it's the continuous monitoring you are implicitly signing up for.
My understanding, from conversations over beers with a few friends retired from the US Navy and NSA:
Signing up for a satphone (or a NEXUS card, or a private pilot's license, or any of a number of other things—including, obviously, secret clearance or access to a security compartment) marks you as a Person Of Interest in PRISM and other monitoring systems. In response, the monitoring systems begin actually caching the event data they see about you into an online data warehouse, rather than just feeding it all into a data lake. The data warehouse is queried out, for each new PoI, to build an individualized ML model (Palantir Intelligence, I think is the software?), where the data initially gathered during your pre-screen is used as the training data—in est, a behavioural baseline. Once the individualized ML model for you is established, it will be run as a batch process (along with thousands of other ML models), about once per day, against any/all newly-arrived PoI event data in the data warehouse. The ML model spits out prediction error; and the system flags anyone with consistently high prediction error as being in need of re-investigation.
In other words, by signing up for a satphone, you're asking the government to keep a detail on you to see if you do anything out-of-character. Just, y'know, a detail that's 99% automated.
People can just drive down a random stretch of desert road near Groom Lake and start catching signals from passing test craft. At some scale you’ve got to give up on the concept of an SCIF and just deal with the fact that people are going to be catching some of what you’re putting out, doing whatever you can to mitigate the damage from there. (Which—if you can at least know where such people are from a satellite view and get guards to them within 10mins of the perimeter breach when they’re still either making their way deeper in, or are driving back out through a bunch of nothing—translates to “you can catch up to them and wreck their equipment.”)
Like I said, I’m not sure how this strategy adapts to the existence of satellite phones. I don’t think it does. So far, instead, satellite phones have adapted themselves to it: you can’t get an Iridium phone without a security screening. I’m not sure what will happen if “private cubesat satphone networks” become a cheap commodity thing to insert into orbit—I think the US and other governments are simply trying to stall that process as long as possible. (Look at the history of satphone networks that competed with Iridium to see what I mean.)