Honestly, I'm neutral about kernel level anticheat. I won't play any of those games, but this doesn't mean we should make porting games that make use of these technologies any harder.
The issue of course is that would reveal just how much of our data those things Hoover up.
That being said, I agree it would be harder to maintain an open source anticheat effectively.
The client can lie about being modified. E.g. it can send false hashes.
How much data a in-kernel or user-land anti-cheat can easily be detected by observing the traffic that flows out from your network, so it really doesn't matter if it's open source or not.
The biggest roadblock to a open source in-kernel anti-cheat is not "exposing the amount of data they extract from you", but rather it exposes how the anti-cheat is working, which is working against the efficiency of the anti-cheat. If you know how they detect you're cheating, it's much easier to overcome that hurdle.
In most cases, security-by-obscurity is obviously flawed, but when it comes to cheat/fraud detection, exposing how it happens makes it's core value less efficient.