← Back

Foundation

Rule Zero

Before you ship anything else, ship trust. The rule that frames every decision after it.

Rule Zero

Don't ship AI you wouldn't be comfortable being on the receiving end of.

That's it. Everything else in this talk, the principles, the eval suites, the governance boards, is just engineering against that one rule.

Why we need a rule above the rules

Standards are great. The Microsoft Responsible AI Standard, the NIST AI Risk Management Framework, the EU AI Act, they're all genuinely useful. But standards describe the floor. Rule Zero describes the posture.

You can comply with every standard on earth and still ship something gross. We've all seen it. The model card was fine. The DPIA was filed. And the product still made someone's day worse.

A quick test

Before your next AI feature ships, ask three questions out loud, in a room, with at least one person who didn't build it:

1. Who is this decision being made about?
2. What happens to them on the worst-case output?
3. Could they tell, appeal, or opt out?

If you can answer those three crisply, you're probably on the right side of Rule Zero. If they make the room uncomfortable, good. That discomfort is the signal.