'Our AI Is Unbiased.' Ask These Questions Anyway.

Sooner or later, an AI vendor will tell you their product is "unbiased" or "fair by design." Smile politely — then ask four questions.

One: What fairness testing did you run, on what data, and can I see the results? Two: Which definition of fairness did you test against — equal outcomes, equal accuracy, or something else? Three: What are the system's known limitations — which groups or cases does it handle worst? Four: How do you monitor for fairness drift after deployment, and will you share those reports with us?

In plain English: "unbiased" is a marketing word. Evidence is a governance word. Ask for the evidence.

A serious vendor will welcome these questions — they've done the work and can show it. A vendor that gets evasive has told you everything you need to know. Either way, you learn something valuable before you sign.

Get the answers in writing, as part of the contract or vendor file. Verbal assurances evaporate; documented answers become your accountability trail. Your fairness obligations don't transfer to the vendor when you buy the tool — they stay with you. So verify like it.

I explain frameworks; I don't give legal counsel.