Five Ethics Questions Before You Launch

You don't need a philosophy degree to run an ethics check on an AI system. You need five honest questions, asked before launch, with the answers written down.

One: Who could this treat differently, and are we comfortable explaining why? Two: What did we train it on, and whose perspective is missing from that data? Three: When it's wrong, who gets hurt — and how badly? Four: Can a human realistically overrule it, with the time and information to do so? Five: If a journalist or regulator asked us to justify this system tomorrow, what would we show them?

In plain English: ethics, operationalized, is just disciplined honesty before you ship.

Run these with the actual team building the thing, not as a compliance ritual after the fact. The point isn't to get five perfect answers — it's to surface the uncomfortable one while you can still change course.

File the answers with your project records. That one-page document is worth more than an ethics principles poster in the lobby. Principles describe who you want to be. Answers describe what you actually decided.

I explain frameworks; I don't give legal counsel.