Pretty much all languages do that. It's a very basic language feature inherited from basic maths notation. Same as x - y subtracts y from x in pretty much any language that supports operators.

Works fine in any language I ever used.

I'm honestly quite surprised that this very basic language feature is even a matter of discussion here.

It's a valid mathematical notation, sure. But there is an implicit understanding that the - in this case is making a number negative rather than subtracting (or, an implicit subtraction from 0).

With the way negative numbers generally work in binary there would be much different ones and zeroes stored behind the scenes, so handling that would have to be pretty intentional.

That said, I did just try it in Java because that's what I work in normally and I swear I had a gotcha with that. But it worked fine as far as I can tell.

Nope, it is not.

x = 5
i = 2
x -= i // x => 3

while

x = 5
i = 2
x = -i // x => -2

x=-i is the unary minus operator which negates the value right of it. It doesn't matter if that value is a literal (-3), a variable (-i) or a function (-f()).

x-=i is short for x = x-i, and here it's a binary subtraction, so x is set to the result of i subtracted from x.

That only works if x is already 0

If i is 10 and x is zero, yes, x -= i would have a value of -10. If x was 5 from something else previously, x-=i would end with an x value of -5.