It's a Unicode bug. Unicode tries to merge different characters into a single code point, because long ago they had the same origin. This particular character exists in Japanese, Chinese, Korean and mathematics, so can be rendered four different ways, but they all share one code point.
Applications have to guess what font to use. Being a mathematical program, this one defaults to the system language (Japanese) but has logic to detect this "no" character and render it in a different font. It isn't clever enough to notice that the rest of the sentence is Japanese, but it shouldn't have to be.
The funny thing is that the same have never been done with latin letters and symbols, because that would be a mess. I really don't understand why they couldn't see it would be the same in Asian langauges.