Skip to content

4 min read

You don't test a car by interviewing the salesman

O'Rume Dominic Uririe gives the AI safety special address at the Lagos AI Summit, Edition I, on 23 November 2026. VOREM, a supporting partner, funds the Proof Award: one hundred thousand naira to whoever tests an AI and finds what the tool missed.

On Monday I am giving the AI safety special address at the Lagos AI Summit. The title is one line, and it is the whole point. You do not test a car by interviewing the salesman.

You would never buy a car by asking the salesman if it is any good. He will tell you it is the safest on the lot. You would still want to drive it, and have someone check it. Because selling you confidence is not the same as proving the thing works.

Right now, across Nigeria and across the world, businesses are doing the opposite with AI. They switch on an agent that talks to their customers and moves their money, and they trust it because the vendor said it was safe. They interviewed the salesman. They never tested the car. That gap is the work I do.

This Monday, in Lagos

The Lagos AI Summit, Edition I, runs on Monday 23 November 2026 at the Oriental Hotel on Victoria Island, from nine in the morning. Its aim, in its own words, is to move Nigeria and Africa from AI experiments to responsible, real deployment. You cannot deploy responsibly on a promise. You deploy on proof. My special address, from London, sits at a quarter past twelve, and that is the message I am bringing to the builders and business owners in the room.

See the full Lagos AI Summit line-up →Lagos AI Summit, Edition I. Monday 23 November 2026, Oriental Hotel, Victoria Island, Lagos.

VOREM is backing it, and putting money behind the message

VOREM, the company I run, is a supporting partner of the summit. And we have done more than lend a name. We funded an award for it: the Proof Award. One hundred thousand naira, to whoever tests an AI and tells us what our own tool got wrong.

Read that again. You do not win by being important. You win by finding the mistake in our work. That is the whole ethos in one prize. We do not tell you our tools are perfect. We pay you to prove where they are not.

Two ways to check an AI, and this covers both

There are two different things to check about any AI. What it does when it runs, and the code behind it. My talk is about the first, the behaviour: does the agent refuse the dangerous action, does it keep a record. The Proof Award is about the second, the code. They are two different checks, and you need both.

The Proof Award uses our free, open tool that reads code and prints what it finds. Three lines, on your own machine:

  • pip install ai-code-quality-auditor
  • cd my-project
  • auditor scan .

Point it at any folder with code in it, yours, not anyone else's. It runs on your laptop and nothing leaves it. Then send the report, plus one paragraph on what it found and what it missed, to dev@vorem.co. Report only. Never send your code.

Entries close on Monday 14 December. The winner is announced on Monday 21 December. It is free, it is open source, and the install count is public on PyPI, so you can check the numbers yourself.

Get the free auditor →Run it on your own code, then send what it found and what it missed to dev@vorem.co

So here is the one thing to take from all of this, whether you are in the room on Monday or not. Test your AI the way you would test a car. Not by asking the people who sold it to you.

O'Rume Dominic Uririe builds verification infrastructure for AI and blockchain systems. The measurement instrument is open source and the receipts library has no dependencies.