I also think you're conflating ASCII with the Latin Script. ASCII is easy to type and read, but anything outside of that range probably presents just as many burdens as say Tamil or Korean. The majority of Latin script languages however, are not limited to ASCII (for example é，ü, or €, which I ironically had to switch to a CJK Input method to type) making your "~70%" closer to whatever the population of native English speakers is.

The "let's make things marginally easy for the majority at the expense robustness in a way where the benefits don't even approach the drawbacks" attitude is how we get things like NodeJS where we have people writing desktop apps in JS b/c the webdevs are more familiar with it at the expense of everyone running massive, slow, and vulnerable apps with 900 different libraries all downloaded from shady sources.

True, you got me, I should have said just ASCII. For the most part though languages use the ASCII set + additional characters rather than replacing or removing characters entirely. so it's still 70%-ish not just English.

I don't see how that's a problem with Arabic specifically. The same would happen if you switched to typing English in Dvorak.