01AI Clean Room
AI works with your data, without leaking it.
The clean room is a conversation mode with the AI assistant built for sensitive work. In this mode, the assistant can read and calculate, but it changes nothing, sends nothing out, and shares only the bare minimum with the language model.

02What the clean room enforces
Six safeguards in every conversation
Read-only, no changes
Only tools explicitly approved for the clean room can run, and they only read data. The assistant will not create, edit, or delete any record.
Contacts stay hidden
Contact email addresses and phone numbers are masked before reaching the language model. The assistant knows the contact exists, but never sends their details to the model.
File structure only
When analyzing a spreadsheet, the model sees column names and row counts, not the cell contents. Charts and totals are filtered strictly by categories, statuses, and dates.
No outgoing email
Sending email is disabled in the clean room. The assistant will not send data to anyone, even if prompted to do so in the conversation.
Approved AI providers only
Your administrator defines which language models are allowed to process sensitive conversations. Unapproved models cannot access them, not even as a fallback during an outage.
Stripped of your identity
Prompts sent to the model exclude the signed-in user's name and the content of the page they are viewing. The model receives only the question and the necessary results.
03Who controls the mode
Turn it on once. It cannot be loosened.
- 1
Active from the first message
You select the clean room when starting a conversation, and it applies to the entire thread. Its restrictions cannot be relaxed later.
- 2
Admins can require it for everyone
Company-wide settings can enforce clean room mode across all conversations, verified on every single message.
- 3
Turns on automatically for sensitive data
For selected modules or record types—such as payroll—the clean room activates automatically. These rules only tighten permissions, never expand them.
- 4
Applies to external AI agents too
An access key for your own external AI agent can be permanently bound to the clean room. That agent will only see the tools permitted inside it.
Good to know
The clean room minimizes data exposure; it is not an encrypted vault. Anything you type directly into the chat will be seen by the approved AI provider. And as always, the assistant only sees what the signed-in user is permitted to see.
True in every mode
- AI never has more permissions than youThe assistant operates with the permissions of the user asking the question. What you are not allowed to see, it cannot see either.
- Changes require review firstOutside the clean room, when the assistant creates or updates records, it proposes the change for your approval by default.
- Every AI action is loggedActions taken by the assistant are recorded in the item's history with a distinct tag, so it is always clear what was done by a person and what was done by AI.
04Frequently asked questions
Frequently asked questions
When should we use the clean room?
Whenever you ask questions involving sensitive data: payroll, HR or health records, contracts, or financial statements. For everyday work where you want the assistant to create tasks or draft emails, use the standard mode.
Can the company enforce the clean room for everyone?
Yes. An administrator can require all conversations to run in the clean room, or enforce it automatically for specific modules and record types.
Will the AI provider see our contacts?
Not in the clean room: contact emails and phone numbers are masked before anything is sent to the model. However, any text you type directly into the conversation will be sent to the model.
Can we choose which AI model is used?
Yes. Your administrator defines the list of approved providers, and sensitive conversations are processed exclusively by them.
05Other security areas
Password Vault Company passwords encrypted in the browser. The server stores them, but cannot read them. Every reveal is logged.
Encrypted Communication Encrypted connections at all times, encrypted emails on request, and end-to-end encrypted direct chat messages.
Operations, Monitoring & Backups EU servers, automatic patches, continuous monitoring, and daily backups with verified recovery.
- Security overview Draivix runs on servers in the EU, every company has its own database, and every change or data export is logged. Here is how the system protects your data automatically.
Let’s start together
Tell us how your business works.
We will show you what a system that takes over part of that work could look like.
Book a consultation