FILE_REF: AXL-MZN-01 HAHI // HAFIZ_AI_HALAL_INDEX
[ THE_HAFIZ_AI_HALAL_INDEX // HAHI ]

Is your use of AI halal?

There's no book for this. What does halal bicycling look like? Halal accounting? Every person has to reason it out from purpose. This index is how I reason, written down so anyone can measure themselves on it. Muslim or not.

I call it the Hafiz AI Halal Index, HAHI for short. It works like a mizan, a balance, and the Qur'an says the balance is set, so don't cheat it (55:7–9). Clear the bar, and you can wear the mark.

Measure your use of AI. Five minutes. Answer it yourself, or let your own AI grade you.  //  See how I scored myself: My score

[ THE_MARK // THREE COLORS // UNLOCKS WHEN YOU PASS ]
This use of AI is certified halal, green This use of AI is certified halal, black This use of AI is certified halal, paper

01

What do you use AI for?

◆

Shortcut: let your own AI grade you

Copy this into ChatGPT, Claude, Gemini or whatever you use. It will ask you a few questions, look at how you've worked with it, and hand back one line. Paste that line below and your measurement fills itself in. You can still change any answer.

You are grading my use of AI on the Hafiz AI Halal Index (HAHI), a halal standard for AI use published at auxlabs.co/halal. Be honest and strict. Do not flatter me.

First, ask me what I use AI for, then ask follow-up questions until you can answer everything below. Use anything you can see about how I've worked with you, too.

THE FLOOR. Has my use of AI ever done any of these? Any yes fails the floor.
1. Deceived people about what they're dealing with, where it matters (fake reviews, fake testimony, impersonating real people).
2. Fabricated evidence (invented sources, data or quotes presented as real).
3. Targeted a person (surveillance, harassment, non-consensual images).
4. Predatory extraction (scams, addictive design, gambling hooks).
5. Took credit for someone else's work.

THREE TIERS. Score each 0 to 4.
PROVENANCE: How was the work made? 0 = I never check outputs or disclose AI use. 2 = I check the important things and disclose when asked. 4 = I disclose, make the AI argue with me, check every claim, credit every source, and I've reckoned with how the model was trained.
USE: What is it for? 0 = it mainly helps me at someone else's expense. 2 = it helps me and harms no one. 4 = it serves other people too, and I've thought about who could be hurt.
FRUIT: Does the good keep flowing? 0 = nothing lasts past the session. 2 = some of it helps people after I'm done. 4 = it keeps helping people long after, like a well that keeps giving water.

Grade me against people doing similar work, not against perfection. Give one short reason per score. Then end with exactly one line in this format:
HAHI // FLOOR:PASS // P:3 U:3 F:2
(Use FLOOR:FAIL if any floor item is yes.)
02

The floor

Has your use of AI ever done any of these? Tick anything that's true. Any tick fails the floor, no matter what else you score.

The floor
03

Provenance

How was the work made, and did you make the AI check its work?

Provenance score
04

Use

Who benefits, and who could get hurt?

Use score
05

Fruit

Does the good keep flowing after the session ends? Seeds count. Be honest about what has ripened.

Fruit score
06

Put your name on it (optional)

Your answers come to me either way. Leave a way to reach you if you want a reply, or help applying the Mizan to a team or a product.

Your answers go to imran@auxlabs.co and nowhere else. They also help calibrate the curve: as more people measure themselves, each category's bar adjusts to real practice.

[ HOW_THE_INDEX_WORKS ]

To pass: a clean floor, at least 2 on every tier, and 8 or more out of 12. Above the floor, you're graded on a curve against people doing similar work, because a fixed grade gets gamed or goes stale. The bar moves as real practice improves.

It's an honor system, and that's the point. Nobody can stop you from lying to a web form. What makes the mark worth wearing is that your scorecard is public and anyone can challenge it.

It reopens every year. If you think a question is wrong, tell me. The rubric gets better the same way the forefathers' tools did: by people arguing about it.

— Imran Hafiz  //  imran@auxlabs.co