Libel suits can be financially catastrophic, so even a tiny false positive rate could present risk that disincentivizes producing such software at all.
And a threat detection mechanism that has a 0.0% false positive rate is conservative to the point of being nearly useless.