---
title: "AI Amnesty Questionnaire"
date: 2026-08-23
description: A structured questionnaire that lets people declare shadow AI use safely, turning an amnesty's honesty into evidence a Board can govern from.
author: Mario Thomas
canonical: https://mariothomas.com/remake/library/ai-amnesty-questionnaire/
---

## What it is {#what-it-is}

An AI amnesty only works if people can tell the truth easily and safely — and if what they tell you arrives in a form the organisation can govern. The questionnaire is that form: the structured instrument the amnesty window runs on.

## What it examines {#what-it-examines}

The shadow AI estate, one declaration at a time: which tools are in use, the specific use cases they support, the types of data being processed, the value already being created, and the risks people have encountered — ten sections designed so nothing critical is missed and no declaration invites blame.

## The readings {#the-readings}

A governable inventory where unknown unknowns used to be: the tools and use cases to sanction, the data exposures to close, the value worth formalising, and the pattern of demand your official provision has been failing to meet.

## Running it {#running-it}

Deployed inside the amnesty window, never outside one: the no-reprisal commitment comes first, the questionnaire second. Declarations arrive per person, per tool; the inventory assembles as they land. Close the window on time and act visibly on what was declared — the amnesty's credibility is the instrument's calibration.

The questionnaire runs in MyMT, my members' workspace — replacing the PDF edition. Deploy it inside your amnesty window and the inventory assembles as declarations arrive.