Not much effect. The Chinese government will prohibit import and possession of the ground terminals. Smugglers will bring in a few but not enough to really bypass the Great Firewall for many people.
Also, even if they were manufactured in the US, China is literally the best country to reverse-engineer electronics and produce knockoffs ;).
That said, AFAIK those satellites aren't meant to be dumb routers/repeaters - so China might just politically/economically pressure US to tell SpaceX to pretty-please don't service connections to/from China.
If you can get a 1Gbps up-link and VPN within China that's going to be harder to track down and could service a lot of people.
However, I suspect SpaceX will however allow China to monitor/limit connections. Further, China has a lot more power to crack down on purely internal operations would deter many.
a) the rooftop CPE needs to be reasonably large and have high gain to be used effectively, and needs clear unobstructed line of sight to as much of the sky and horizon as possible.
and
b) It needs to transmit, it will transmit in very known frequency ranges, and the chinese authorities have plenty of budget to buy portable spectrum analyzers, horn antennas and to reach people how to use them. It only takes a couple of hours maximum to reach somebody how to use a portable spectrum analyzer to physically locate and identify transmitters in almost any band.
A) all radome materials have some degree of loss. Ideally you want no loss at all. The path loss in higher than 10GHz frequencies to LEO is already extreme.
B) not just up but to the sides and toward the horizon as well. This will be a non moving phased array antenna that can talk to two LEO satellites moving across the sky at the same time. I predict that any reasonable amount of blockage to the sides will not be a good idea for their network architecture.
C) all Tx have some sort of sidelobes and nothing has a perfect f/b ratio. Will still be detectable by spectrum analyzers from the side.
A) 1% signal attenuation is not going to kill this and that's plenty to work with.
B) Depends on how many satellites are in the constellation. Ideally you want a lot of them as there is vastly more atmosphere the lower your angle to the horizon.
C) detectable at say 50 feet sure, but a signal you can detect from 10+ miles in a van doing 60mph without a lot of false positives is another story.
the threat model for hiding from spectrum analyzers is not 10+ miles in a van doing 60 mph. With the resources of a nation-state at the disposal of the spectrum analyzer operators (Iran, China, Ethiopia) it will look more like a person walking around a neighborhood with a portable spectrum analyzer and horn antenna. Multiply by however many people are needed to canvass a metro area in a reasonable amount of time.
With the resources of China, spending $15,000 per spectrum analyzer kit x 8 kits, plus training, is a tiny drop in the bucket compared to the amount of money and effort they are currently spending on IP cameras that feed into facial recognition systems, etc.
I was more thinking a FSO link ~2km into the middle of nowhere vs having one of those sitting in someone's house. But directional wifi is probably safer as it's also not going to look strange to a spectrum analyzer.
However, if you want to catch individual usage in urban areas, with guys on for that's going to easily be into the 10's of of billions as users don't need to keep them on 24/7. AKA within the capability of nation states, but not without significant effort.
And those people who would be interested in smuggled ground terminals will already be using VPNs. Maybe this will be more convienient and possibly somewhat safer, but not it won't be a game-changer for even a minority (at least not in the firewall problem, general internet connectivity in rural China is another issue)
