Open firmware, but who says the silicon isn't backdoored? And why is open firmware more important for your freedom than open silicon? What about on-chip ROMs? :-)
In the end, you're always drawing arbitrary lines in the sand. If you really want to go all the way to actually reach a tangible goal, I'm only aware of one project that actually offers trust guarantees backed by hard engineering arguments: Precursor
https://www.crowdsupply.com/sutajio-kosagi/precursor
(TL;DR on the trick is that, by using an FPGA, you make it nearly impossible to backdoor, because it would take a huge amount of compute power to engineer a backdoor into the FPGA silicon that can analyze arbitrary randomized FPGA designs to backdoor them).
For more practical computing, I find M1s have a very solid secureboot design. We can retain that security even after putting m1n1/Asahi on there; doing so requires physical presence assertion and admin credentials and locks you into a single m1n1 version without repeating the process, so we can use that to root our own chain of trust. Similarly we can continue to use Apple's SEP (in the same way one would use a TPM, HSM, or a YubiKey for that matter; sure, it's a proprietary black box, but it also can't hurt you except where you let it) just like macOS does, for things like encryption and storing SSH secrets. And all the coprocessors have IOMMUs, and the whole thing is very logically put together (unlike the giant mess that are Intel CPUs; e.g. we know they have a JTAG backdoor in production chips after authenticating with keys only they have, never mind things like ME) and designed such that the main OS does not have to trust all the auxiliary firmwares.
I'd love to have a more open system competing in the same space, but I have no problem trusting Apple's machines much more than I do the x86 PCs I use these days anyway, and it's hard to find things that compete in the same performance/power space that are more open than either of those, unfortunately.