Use case · By risk
Keep sensitive data out of the wrong AI.
Sensitive values are detected and masked in the path, and logged.
“Personal and confidential data gets pasted into AI.”
Control in three steps
Sensitive values are detected and masked in the path, and logged.
Check who is asking
Every person and agent is named, with a role from your directory.
Allow, ask or decline
Inside the role it runs. Changes wait. Outside the role, declined.
Record it
Every step lands in the audit trail.
See it work
Watch it happen.
Step 1 of 4: Request
“Draft a reply to the member with ID 28XXXXXX about her claim.”
Every request starts with a named person and a named agent.
Simplified example. Names and records are invented.
Text version
Mercy A. (Customer Service) asks Staff assistant: "Draft a reply to the member with ID 28XXXXXX about her claim." The security check returns: Masked. Every step is written to the audit log.
01 Request
Mercy A. · Customer Service · Staff assistant
Draft a reply to the member with ID 28XXXXXX about her claim.
02 Security check
Masked
- Check
- Sensitive data scan
- Found
- National ID number
- Action
- Masked before it reached the model
The AI sees a placeholder. The real value never leaves.
03 Human approval
Not needed
The request was handled at the check. Nothing waited for a person.
04 Audit log
- 09:14:02 · request · mercy.a · staff-assistant · prompt
- 09:14:02 · check · sensitive value · id number
- 09:14:02 · masked · masked: 1 · prompt sent
- 09:14:02 · logged · event 7f3a…e21 · on the record
Questions
Short answers. Ask us anything else on the call.
See this in your own organisation.
Start with a free assessment. Two weeks. Read-only.
A one-page answer for your CEO and CISO. Assessments are offered at our discretion.