You could go full air-gapped and get a machine without any network cards. Would still recommend a nice bunker deep inside a mountain just to be on the safe side.
The implication (intentionally or not) being that we can just as well continue to run Chrome on Windows 11 on open WiFis because everything is broken all the time and if not any Three Letter Agency can always kidnap you or someone you care about and get their secrets that way.
Encryption works and massively increases the workload for would be drag net operators!
Having a choice of technology is good and also increases the workload for would be drag net operators.
Competition is good and forces companies to get their act together.
Trashing every good thing that happens because it isn't perfect is demotivating and I'm also absolutely fed up with it.
Every respondent so far assumed you knew this (because it's an unusually high-knowledge question to ask and an unusually obvious answer) and so assumed the actual purpose of your post was to declare that you personally wouldn't use an environment where you can't be sure the SoC isn't doing something nefarious. The people replying don't really find that declaration interesting. Your subsequent reply says that the choice to live with this particular risk is some sort of tacit acceptance, which implies you are not tacitly accepting it.
Since no such environment exists, except for edge case hypotheticals like only powering an air gapped computer inside a faraday cage in a lead bunker, most people expect that you too are tacitly accepting it, which makes this particular complaint feel like a non sequitur and not an invitation to discuss the broader issue.
If you have an actual real life use case about trying to harden a system to this degree, like you are responsible for nuclear safety or the unlock codes for the secret lunar base or something, I am sure people would like to read about it.
There's a reason China and Russia build their own CPUs. There's nothing else you can do to ensure that nothing leaks.
In practice, you could probably run the machine for a while hooked up to an ethernet cable and sniff the outgoing packets. I don't think the M1 will send Apple telemetry on its own (though it certainly could, if it wanted to).
I also guess more advanced adversaries will add stuff that only becomes active after a long time, and may use steganography to hide what it’s sending.
So, bootstrap your system the way this guy does: http://paillard.claude.free.fr/video.mp4
(I don’t think he has gone past vacuum tubes)
I wonder how the Chinese and Russians have validated custom chip designs. I don't think the compiler would be bugged specifically for chip design software, but there's no real way to tell...
https://www.crowdsupply.com/sutajio-kosagi/precursor
If you're willing to trust that the silicon isn't secretly backdoored in some ridiculous invisible way, well, unlike Intel, the M1 does not have any hyper-privileged proprietary firmware by all indications, and that proprietary firmware which does exist is plaintext and analyzable (and not privileged to scan memory), so I have a much easier time trusting these machines than a random x86 PC (all-privileged SMM supervisor and ME/PSP) or Android phone (all-privileged TrustZone supervisor).
But if Apple were really clever they'd add a secret undocumented radio built into the SoC package that sends signals modulated in an unusual way over an unusual low-bandwidth but high signal frequency (probably sub-CB Radio) that no one is monitoring.
Which now brings you to the scenario that you have to use a machine that not only never uses a network, but also exists in a physical place that no network signal could ever penetrate. And it goes without saying that having taken this level of caution, one should probably buy the computer with a fake name and all cash, ideally through a network of international proxy buyers, to be sure.
This is actually rather trivial to achieve, just put it in a faraday cage.