AI Workspace for Sensitive Data.
Review the contract between ⟨PERSON_1⟩ and ⟨ORG_1⟩.
Pay the invoice to ⟨IBAN_1⟩ by Friday.
Send the signed mandate to ⟨EMAIL_1⟩ today.
Call the client on ⟨PHONE_1⟩ before the closing.
Add the AHV number ⟨AHV_1⟩ to the pension file.
Client data can’t go into consumer AI tools. So most confidential work never touches AI.
Professional secrecy at risk
Consumer AI tools were never designed for Art. 321, § 203 or Art. 226-13 obligations.
Data used for training
Free chatbots may retain and train on whatever your team pastes into them.
Zero process visibility
No record of what left the firm, which model received it, or under which rule.
Every leading model, behind privacy infrastructure built in Switzerland
Mask client data before it leaves. Names, addresses and case references are replaced before the request leaves, and restored in your answer.
Every model, one subscription. GPT, Claude, Gemini and Mistral on one invoice. Auto picks the best model for each task.
AI that works with you. Hand Cedreon a piece of work, not only a question. It reads the file, checks the sources and drafts the summary, behind the same vault.
One account for the whole firm, governed by your rules
Custom Guardrails
The firm’s rules apply to every member. Anyone can add their own on top, and no one can loosen them.
Regions locked
One switch restricts the whole workspace to models built and run in Europe.
Check the list, then send
Attach a contract and Cedreon lists every value it will hide. Add any it missed, then send.
Organisation account
One admin console for the whole firm: add members by link, set the rules and the models once, and share the documents everyone works from. Conversations stay each member’s own.
Seats on one invoice
Buy seats once. Invite with a personal link; access activates on first sign-in. No IT project, no mail server.
Secrecy agreements, countersigned
Professional secrecy is national law, so we sign it with you where you practise: CH, DE, FR and more across Europe.
Which AI models can I use?
GPT, Claude, Gemini and Mistral, on one subscription and one invoice. Choose per conversation, or let Cedreon choose for you.
What exactly gets masked?
Names, organisations, IBANs, card and client numbers, addresses and case references, plus anything your own rules add. Values are replaced before the request leaves us and restored in your answer. Detection is layered and best-effort: checksum patterns first, then a name recogniser that runs on our own servers.
Is our data used for training?
No. Every request carries an explicit no-training instruction, and the obligation is contractually binding on each provider we route to. How long a provider holds a request in its own operational logs is a separate question and varies by model.
Where is our data stored?
Your conversations, documents and audit chain are stored encrypted by Cedreon and never leave it in the clear. For the model itself, an organisation can lock its workspace to models built and processed in the EU.
How does billing work?
Seats, once, on one invoice for the whole firm. Every model is included, and new ones join your subscription as they ship. You start on a free allowance.
Can we see what the model actually received?
Yes. Every exchange lands on an append-only ledger: what was sent, what was hidden, and by which rule.
How do we get started?
Create a free account and confirm your email address, then try it on a real matter. When you are ready, your admin buys seats and hands out access with a personal invitation link.
Ready to try it on a real matter?
Join firms across Switzerland and Europe using Cedreon to bring AI to confidential work. Free to start.