SAFE-DOC · PSEUDONYMIZATION
Your teams use AI. You keep your secrets.
Safe-Doc is your AI firewall that replaces personal data with aliases before sending to Claude, ChatGPT, Mistral, Gemini or Copilot, then restores the real values in the response, locally, on your side.
10 days free. No commitment, no credit card.
Before / after comparison
Drag to compare
Original document
Protected document
orange = personal data bleu = Safe-Doc token
- ClClaude
- GPTChatGPT
- MiMistral
- GeGemini
- CoCopilot
- CuCursor
- PxPerplexity
- CwCowork
- ViVibe
- GrGrok
Launch
10 days free. No credit card.
Full access now. Pseudonymize real documents, keep the AI of your choice. No card required to start.
ZDR (Zero Data Retention): nothing is retained with us, not documents, not queries, not content history.
The problem
Banning AI doesn't work : it's already here.
Your teams already paste confidential documents into ChatGPT. Figures below come from our Shadow AI page : cited sources, not extrapolated.
Sources: IBM, Cyberhaven, Software AG : details on AI Risks
Positioning
A security layer, not another platform.
A security layer, not another platform. "Workspace + integrated chat" platforms store your files and see your prompts: versions, history and mapping tables pile up with a third party. Safe-Doc processes, restores, and hands control back. Your file never becomes an archive on our side.
Two uses
Two ways to use Safe-doc
Safe-doc does one thing, pseudonymize before AI, but you can use it in two ways: day to day, to protect a prompt or a document, and in your processes, to hand AI whole tasks on confidential data.
Day to day: protect a prompt or a document
From the Safe-doc app, you drop in a document or paste a text, review what will be masked, then send it to ChatGPT, Claude, Copilot or Mistral. The answer comes back with the real names, on your computer. No document is kept.
If you would rather stay in the chat window, the optional Chrome extension does the same thing directly in ChatGPT, Claude, Gemini, Mistral, Perplexity and Copilot: your message goes out protected, and the answer shows the real names.
In your processes: hand whole tasks to AI
Payment reminders, bank reconciliations, summaries, client onboarding: with the Claude Cowork skill, the API, the MCP server or x402 pay-per-call, your AI agents handle complete processes on personal data. They work on aliases, and the real data only comes back to you, at the end.
Interim tool
The group project takes months. AI usage is already here.
Safe-Doc doesn't wait for your enterprise platform: it's the layer that secures existing usage during the transition : no lock-in, no conflict with the roadmap ahead.
Objection : "We're waiting for the group IT solution."
Safe-Doc deploys in MINUTES and secures existing usage while the group project progresses, and stays useful afterwards, whatever AI is chosen.
The three freedoms
Keep your tools. Add the safety.
Tool freedom
Any AI: Claude, ChatGPT, Mistral, Gemini, Copilot. No lock-in, no quota. Your subscription stays yours, and it's always the best model available.
Workflow freedom
No workflow change, no training. Protection is added; your teams change nothing. That's what makes it adopted instead of bypassed.
Vendor freedom
ZDR. Nothing stays with us: nothing to audit elsewhere, nothing to retrieve if you leave, no lock-in.
Keep your AI and your subscription. No lock-in, no quota.
Beyond risk
Speed, cost, always the best AI.
Speed
Live in minutes. No migration, no platform to learn, no hardware requirement.
Cost that follows adoption
Start at €19, grow in seats and pages at the real pace of usage. No upfront platform commitment to justify before proving value.
Always the best AI
A dedicated AI is frozen on one model; the frontier moves every quarter. Safe-Doc stays a model-agnostic layer: always plug in the best model, today and tomorrow.
By role
The same truth, for your role.
How it works
Pseudonymise → analyse → restore.
Same AI tools. Web UI, API or MCP. No sensitive data leaves your control.
90+ entity types replaced with neutral tokens. Metadata wiped. Single document or Data Room batch with consistent pseudonyms.
Copy-paste, call via API, or connect through MCP - into Claude, ChatGPT, Mistral, Gemini, or Copilot, your choice. Safe-Doc never sees your query. No integrated chat. No lock-in.
Re-inject original data into the AI output via encrypted mapping. Locally. Auditable. Under your control only.
Who uses Safe-Doc
Built for anyone handling sensitive documents.
Not just lawyers. Any professional team that uses AI on confidential content.
Client files, NDAs and legal packs, pseudonymised before any AI review. Aligned with professional secrecy requirements.
Deal documents and data rooms with consistent pseudonyms across multi-document batches.
Client deliverables secured before AI-assisted analysis or drafting. No sensitive data exposed to third-party models.
Financial statements, audit files and sensitive datasets, without exposing figures or counterparty names.
Employee files and HR procedures processed with AI while staying GDPR-compliant.
Enable AI adoption across your teams without compliance risk. One security layer. Any AI tool.
Architecture
ZDR · Zero storage by design.
Temporary processing. No document content retained in logs or database after treatment.
Encryption in transit. Only the pseudonymised content reaches the AI of your choice.
Operation logs and residual risk indicators included in every export report.
Pricing
Planned launch pricing
10 days free on all paid plans. No credit card to start.
- 1 seat
- 30 pages / month
- FR · EN · DE anonymization
- N1 and N2 levels
- PDF, DOCX, TXT, emails
- 1 seat
- 1,200 pages / month
- Everything in Free
- Zero-persistence mode
- Priority support
- 1 seat
- Everything in Solo
- API access
- Integrations (Claude Cowork…)
- 3 seats
- 6,400 pages / month
- Everything in Solo
- +€35 per extra seat
- 10 seats
- 20,000 pages / month
- Everything in Cabinet
- +€35 per extra seat
- Unlimited seats
- Unlimited pages
- SSO
- Audit logs
- Dedicated infrastructure
- DPA & SLA
Seats
- Extra seat: €35 (Cabinet and Cabinet+).
Page packs
Volume packs (all paid plans):
| Pack | Pages | Price | Per page |
|---|---|---|---|
| S | 500 | €15 | 3.0¢ |
| M | 2,000 | €45 | 2.25¢ |
| L | 5,000 | €99 | 2.0¢ |
You or your teams already use AI on confidential documents : with or without your approval.
That's Shadow AI.
We don't ban AI. We make it safe. GDPR art. 4(5) pseudonymisation : a recognised security measure, not a substitute for your governance.