Although I understand the rationale, lumping vague probability words like "may", "could", "might" as antagonyms (auto-antonyms) doesn't seem valid.

Whether you use "may" or "may not" would indicate the leanings of your null hypothesis / default assumption.