Take Inventory of Your Models

Grants bought equipment with federal money, and every item went on an inventory: what it is, where it is, what it's for, what condition it's in. Once a year, someone physically verified it all. It seemed fussy until the audit asked "where's the laptop?" and we could answer in thirty seconds.

Now do your AI.

Most organizations can't answer the equivalent questions: Which AI tools are we actually using — including the ones employees signed up for on their own? What data does each one touch? Who's responsible for it? When was it last reviewed?

That's your model inventory, and building it is the same exercise:

1. List everything. Every AI tool, model, and feature in use. Yes, including shadow IT.
2. Note what it touches. What data goes in, what decisions come out.
3. Name an owner. Every item needs a human responsible for it.
4. Review on a schedule. Inventories rot. Re-verify regularly.

You can't govern what you can't see. Grants knew that about equipment. It's just as true about algorithms.

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