Model Cards in Plain English

A model card is a nutrition label for an AI model. Just like a food label tells you the ingredients and what's inside, a model card tells you what an AI system is made of and what it's supposed to do.

A good model card answers plain questions. What is this model for? What data did it learn from? How well does it perform — and on whom? Where does it struggle or make mistakes? Who built it, and who do I contact with questions?

Researchers proposed the idea a few years ago because AI systems were being shared and reused with no instructions attached — like handing someone medicine with the label torn off. The card fixes that.

Here's the practical part: you don't need to build models to use model cards. If your team uses an AI tool — a hiring screener, a customer service bot, a forecasting model — write a one-page card for it. What do you use it for? What have you noticed it gets wrong? Who's responsible for it?

One page. Plain language. Updated when things change. That single page will answer half the questions any auditor, client, or curious executive ever asks you.

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