Ham radio in the US has restrictions on sending encrypted data over ham frequencies.

They can't stop you, they probably won't even catch you unless you are egregious about it. But if they do catch you, it's like a $10,000 fine.

I think the whole world is quickly moving closer to China's model. Everything that gets encrypted will need the government's key on it anything they can't decrypt will get blocked.

You basically set up some rules at the backbone level looking for suspect traffic. They could now have AI review the suspect traffic and try to tell if what's going on is viable data or nonsense words/coded messages. All communications will need to be identified. None of the blocking would work real time but once they know who's sending it in think that you've sent some stuff that you shouldn't be sending they could just turn you off.

I read an article somewhere recently where AI was able to tell if an image was being used with even the most advanced steganography with a fairly high reliability.

They'll never be able to stop people from privately communicating at small scale, But man will there be some watch lists.