Blog · Security

Where are your AI keys right now? A five-step check

Every AI tool your team builds with runs on an API key. And those keys have a habit of ending up in strange places: a script on someone's laptop, a message in Slack, a shared doc, or committed straight into the code.

That matters for two reasons. Anyone who finds a key can run up your bill. And a key in the wrong place is hard to switch off without breaking whatever depends on it.

The good news: tightening this up usually takes an afternoon.

1. Find the keys you already have

Search your code repositories, shared drives and team chat for anything that looks like a key. Ask each team which AI providers they use, and check the provider dashboards for keys nobody recognises.

2. Move them into a vault

Put keys in a secrets manager or your cloud provider's key vault, and have apps read them from there. Keys should never live in code, documents or messages.

3. One key per team or app

A single shared key means you can't see who is spending what, and you can't revoke one without breaking everything. Give each team or application its own key.

4. Put a spend limit on every key

Most providers let you set monthly limits and alerts. Set them. It turns a nasty surprise into an email.

5. Rotate anything that has leaked

If a key has ever been pasted somewhere it shouldn't be, assume someone has seen it. Create a new one, update the apps, and delete the old one. Removing it from the message or the code isn't enough; copies live on in history.

Make it stick

Add one line to your AI use policy: keys live in the vault, nowhere else. Then check again whenever a new tool or team starts using AI.


Want a second pair of eyes on how your firm uses AI? Book a free 20-minute call.

More from the blog

Let's talk about what you're building.

A free 20-minute call. No sales pitch.

Book a free call