What should you ask a private AI vendor?

Ask twelve things: where data sits, which model reads it, who sees what, what the log shows, who runs it, what it costs and how you leave. A vendor that cannot answer each one plainly is not ready for your data.

Last updated 11 November 2026 · Mert Uzun, Co-Founder

Why it matters

1 in 5

organizations was breached through shadow AI. (IBM, Cost of a Data Breach 2025)

97%

of organizations that had an AI-related breach lacked proper AI access controls. (IBM, Cost of a Data Breach 2025)

How it works in practice

1. Where is our data stored, and in whose account?

2. Which model reads our content, and can we bring our own?

3. Can it run on private models in our own cloud or on-prem?

4. Is restricted content filtered before the model sees it?

5. Does it check each person's access, or does one key see everything?

6. Can we read the log of who asked what?

7. Who installs, updates and monitors it?

8. What hardware does it need, and who buys it?

9. What does it cost beyond the licence?

10. Which of our tools does it read, and how fast is a new one added?

11. Can the agents our people already use ask it?

12. What happens to our data if we leave?

Ask for answers in writing, and ask to see each one working on a real pilot.

How Mithrandir does it

  • Runs in your own cloud on AWS, Azure or GCP. We never host your data.

  • Can run with private models in your own cloud or on-prem. Talk to us.

  • Works inside the agents your people already use.

Questions

Which questions matter most?
Where data sits, and whether each person's access is checked. Everything else builds on those two.
Should we ask for a pilot?
Yes. One team, one set of tools, real questions.
Where can we compare vendors?
See the private AI comparison pages on this site, each with sources and dates.

Remember.

Every company needs agents that know it. We are here for that.

Remember more. Spend less.

Bring one workflow. See it on your own data in 30 minutes.