Personal AI
Policy Maker

A Brief Guide

Clients, employers, and boards are starting to ask where you stand on AI. The Personal AI Policy Maker runs an interview with Natalia, an AI who asks about how you actually work and turns your answers into a written policy — your red lines, what you protect, when you disclose AI's role, and what you do when it gets something wrong.

It reflects decisions you have made, in your own words, not a generic responsible-AI checklist.

How the Interview Works

The interview runs one question at a time and adjusts to what you tell it. Nothing about it is a form to fill out.

Natalia interviews you

One question at a time, about how you actually work. No model answers, no ethics lecture. She tests an answer that sounds like the one you think you're supposed to give.

Answer by talking, typing, or adding a file — she reads uploaded policies, requirements, or notes before she asks anything.
The question you need to answer is always in bold. Everything else is her reflecting back what she just heard.

You see the policy take shape

About every three or four topics, a short checkpoint tells you what she thinks she's learned — a boundary you set, a tension still open. No surprises at the end.

You choose the format and approve it

One page, practical, or comprehensive. She drafts it in first person from decisions you actually made, runs a final stress test, and revises once more before you approve.

Approve any time it's genuinely done. You are not required to run the full stress test if you already have what you need.

Start Your Policy Interview

What Your Policy Covers

Ten sections, adapted to the length you choose.

My North Star. Your purpose, values, and stakeholder commitments.
Boundaries and Red Lines. Where and how you will and will not use AI.
Data, Confidentiality, and Research. What you protect, the minimum-necessary standard, and where research becomes profiling.
Disclosure and Provenance. When AI involvement is material enough to say so, and how.
Human Review and Decision Rights. Where judgment, verification, and authority stay human.
Error, Remedy, and Accountability. What you do when AI-assisted work causes an error or harm.
Bias, Inclusion, and Constituent Voice. How you examine AI-enabled work for unequal impact and missing perspectives.
Intellectual Property. Your standard for authorship, attribution, and community-held knowledge.
Procurement and Recommendations. The diligence you owe before recommending, piloting, or buying AI for someone else.
Responsible Experimentation. How you tell experimentation from deployment, and when to stop or redesign.

It closes with 8 to 12 concrete decision rules — if this situation, then I will do this — plus a table of anything that needs outside verification, kept separate from what you decided yourself.

Put the Policy to Use

Hand it to whoever is asking. A client, an employer, a board, or your own team, at whatever length fits the audience.
Keep one canonical copy. Somewhere you keep your working documents, not scattered across drafts.
Reread it when something changes. A new client, a new tool, a new kind of AI use — revisit the policy rather than assume it still holds.
Bring it to a conversation, not just a drawer. The value is in using it to make a real call, not filing it away.

Bring It to CREATE

The Human-Centered AI Conference

October 16–18, 2026 · Salt Lake City

Gather With Us