That's not actionable. The dark part here is the serious threat of applying oppressive prior restraint policies at will. Employing machine learning in the process would only offer a false sense of legitimacy.