Is negative testing only non-functional?

In the context of software testing, negative testing means testing conditions that are not specified (e.g. putting an alphanumeric char into a numeric field, processing a file that is not supported...). But when I think about that, all of this is actually non-functional testing evaluating the robustness of the system, reliability of recoverability. Or am I missing something?

I assume this is really the case as with negative testing, I answer the question "is the system robust enough"? Which is a typical example of non-functional requirement.