>How would that even work? The adblocking process has full access to whatever's being passed through it, and therefore it can eavesdrop or tamper (eg. inject malicious js) with the contents.
A web browser is so much more than just the network traffic. Consider the mouse & keyboard input, the browsing history, the cache, the security credential store, the connection to display, GPU, USB and Bluetooth devices, microphones & cameras... the list goes on. This includes ability to read and write local files, start new processes, compile and run code. And then there are other HTTP user agents, like automatic software update tools.
Most of this functionality shouldn't be needed by, nor accessible to, an ad blocker.
Lastly, you could put much stricter resource usage limits (CPU time, operating memory) on an adblocking process than on a web browser, thwarting whole classes of resource exhausting attacks.
>only worse because it only works locally
You can apply varying levels of restricting any net traffic not passing through the adblock, up to and including dropping any such traffic. Ability to ad-filter HTTP traffic of an email client, RSS reader (i know, 2020) and similar user agents would be good.