What _exactly_ that means is, of course, up for debate. Some say you can use encryption as long as the key is publicly available and obvious (i.e. my key is "password"). There's been a bunch of stuff happening in the last year around Winlink. Devil's in the details.
We both have a shared key, I send you a cat picture over ham radio with the message encoded in the least significant bit of each pixel.
How exactly is that going to be detected? The picture looks virtually the same as the original.
Not everything is about whether or not you get caught.
If they claim is not encrypted it's just noise, prosecute them for broadcasting noise.
If they claim it's not noise it's just encrypted, prosecute them for broadcasting an encrypted signal.
What do you think the practical difference in outcome is?
D-Star and PACTOR are proprietary, undocumented algorithms but you're allowed to use them because devices implementing them are commercially available.
What's the advantage of doing it?
Mainly a mitigation against curious people with Ubiquiti equipment, which makes it way too easy to disable country restrictions and find and connect to our links in licensed bands :-)
Alternatively there might be protocols which have mandatory encryption as part of their specification. Or you might be using a combined encryption-authentication mode, but only for the purpose of authentication on ham radio.
Those are all a bit of a stretch though.