> Except that is not AI safety.

Your assumption is uninformed and patently wrong. One of the primary concerns of AI safety is surveillance.