Text this: AI, Trustworthiness, and the Digital Dirty Harry Problem