
Data Privacy & Residency
Your data stays where your contract says
Most AI rollouts start by sending data out. You choose the vendor, model and region, and the contract decides.
- Vendor, model and region chosen against your contract
- Self-hosted weights where the contract requires it
- Personal data refused before it reaches a prompt
- One process, whichever option you choose
Example record
What your boundary caught
Each prompt is checked before it leaves your network. You see what went out and what stayed.
Each refusal is logged with the rule that caught it.
- Sent
Support ticket summary
no personal data found
- Refused
Ticket containing a card number
credential-shaped string
- Refused
Call to an unpinned region
outside the contract region
- Sent
Batch to the self-hosted model
stays on your infrastructure
Where data leaves without anyone deciding
Data rarely leaves through a decision. It leaves through a default nobody revisited.
A provider default that retains prompts
A test environment wired to a production key
A document pasted into context with a customer name in it
A region chosen by whoever set up the account
Capabilities
Where your boundary sits
So your security team reviews it like code, in your repository.
Model
What we fix, and what you decide
What we fix
- F1Data rules written down before a prompt is sent
- F2Checks on each prompt, configured in your repository
- F3Refusals logged with their reason
What you decide
- C1Vendor, model and region
- C2Hosted API or self-hosted weights
- C3Which data classes may enter a prompt
Related
Where to go next
Next step
Tell us where your data has to stay.
We map what leaves today. You get a list of changes your legal team can check against the contract.

