Anonymize and pseudonymize client documents without sending them to the cloud.

Celatum detects and anonymizes personal data in Word documents, entirely on your infrastructure. Built for lawyers, fiduciaries, and healthcare professionals bound by professional secrecy.

✓ 100% offline · ✓ DE / FR / IT / EN · ✓ Aligned with FADP & GDPR

mandate.docx
Dear Person A,
Your mandate with Law Firm 1 was updated.
Contact:
AHV: 756.XXXX.XXXX.XX
4 identifiers detected, anonymized locally

Features

Professional anonymization of personal identifiers for legal, compliance, and enterprise teams.

AI-Powered Detection

Our AI model detects 50+ personal identifier types — names, emails, SSNs, addresses, phone numbers, IBANs, dates of birth, and more. Runs on CPU, no GPU required.

Pseudonymization

Consistent, document-scoped replacements (Person A, Organization 1) preserve readability while masking real identities. Export pseudonym key files for reversible workflows.

DOCX Support

Drop files or paste text. Process single files or batch entire folders. Output to DOCX with anonymized or pseudonymized content, ready for filing or publication.

100% Offline

The AI model runs entirely on your local machine. Documents never leave your device — no cloud, no uploads, no third-party APIs. Air-gapped environments supported.

Review & Control

Review every detected identifier before anything changes. Enable or disable individual findings, add your own terms, and adjust labels. You decide exactly what gets anonymized.

De-pseudonymization

Restore the original identifiers using encrypted key files. Reversible anonymization for internal review, legal proceedings, or audit trails.

Multilingual Glossary

Built-in glossary with Swiss legal terms in EN, DE, FR, and IT. Translate personal identifiers labels, titles, and legal terminology. Fully customizable per user.

Privacy by Design

Built to ensure your processing stays compliant. Documents never leave your device: no transfer, no third-party processor, no cross-border data flow to assess. Secure memory zeroing, AES-256 encrypted license data, append-only audit log, no telemetry.

Reusable Templates

Save pseudonymized documents as .piia templates. Re-populate with real or alternative data later — ideal for court filings, case studies, and recurring document workflows.

Security by Design

Security from day one. No data ever leaves your infrastructure.

Hardened by Design

Encryption for stored keys, OS-level access controls, automatic memory wipe at session end, signed and timestamped builds, and a minimal network surface. Every layer engineered to fail closed.

Confidentiality First

Document content never leaves your infrastructure: no uploads, no telemetry, no error reports. License and update checks carry zero document data.

Professional Secrecy

Designed for professionals bound by legal confidentiality obligations. Compliant with GDPR and the Swiss Federal Act on Data Protection (FADP/DSG). All personal identifiers processing is fully local: zero data transmission to third parties.

Independent security statement

Celatum’s security has been independently assessed: following a penetration test of the application conducted on 16–17 June 2026, OneConsult AG, a cybersecurity firm based in Zurich, issued an independent security statement on 3 July 2026.

About us

Celatum was born from the difficulty small and mid-sized law firms in Ticino face in using artificial intelligence without compromising data protection and professional secrecy. Tools like Claude, ChatGPT, or Gemini have become decisive for staying competitive, but sending briefs containing personal data to cloud services is hard to reconcile with the FADP and professional secrecy.

Our application anonymizes documents directly on your device (no content ever leaves the computer), removing personal data before the text is used with a cloud LLM. That way even a small firm can leverage AI on its own documentation, keeping control of sensitive data, at an accessible cost.

ESEdoardo Salvioni
Edoardo Salvioni

Founder

Developer with experience in AI and blockchain.

LinkedIn
RSRiccardo Salvioni
Riccardo Salvioni

Co-founder

MLaw | Attorney-at-Law | Financial services & digital law consultant | Legal AI enthusiast

LinkedIn

Pricing

Simple, transparent pricing. All features included. No hidden costs.

Professional

CHF 50 /seat/month

Billed today: CHF 50

Volume discounts applied automatically at checkout.

Cancel anytime: takes effect at the end of the current billing period.

  • AI-powered detection of personal identifiers (50+ types)
  • DOCX anonymization with formatting preserved
  • Pseudonymization & de-pseudonymization
  • Reusable .piia templates
  • Bulk anonymization
  • Multilingual glossary (EN, DE, FR, IT)
  • 100% offline: no data leaves your infrastructure
  • Secure memory zeroing after processing

Multi-seat discounts: lower per-seat rate from 6, 11 and 26 seats

Need more than 50 seats or custom deployment? Contact us for enterprise pricing.

Frequently asked questions

Is Celatum compliant with the Swiss Federal Act on Data Protection (FADP)?

Yes. Celatum is designed around the local-processing principle: documents are processed directly on the user’s device, and their content is never transmitted, stored, or logged on external servers. This approach reduces the risk of data exposure and supports compliance with the principles of data minimisation, purpose limitation, and security of processing under the Federal Act on Data Protection (in particular Art. 6 and 8 FADP), as well as the corresponding GDPR principles, including Art. 5 and 32.

How does Celatum compare to cloud LLMs for anonymization?

Cloud LLMs normally require sending the original document content to third-party servers. When the document contains personal data, confidential information, or material covered by professional secrecy, this step can carry significant risks under the FADP and, for certain professions, also with respect to the professional secrecy obligations of Art. 321 of the Swiss Criminal Code. Celatum takes a different approach: personal identifier detection happens locally, without the document ever having to leave the user’s device or infrastructure. In practical terms, cloud LLMs can be useful for drafting, summarising, or rephrasing non-sensitive content. Celatum, by contrast, is built for anonymizing confidential documents, personal data, and client material.

How is this different from manual redaction tools?

Manual redaction tools require you to select each item to anonymize, one at a time. Celatum automatically detects names, addresses, AVS numbers, IBANs, dates of birth, and 45+ other identifier types in DOCX files, lets you enable or disable them in a review panel, and applies the anonymizations in a single pass.

Can I use one seat on more than one device?

Each seat activates one user on one device. For firms, multi-seat plans group several seats under a single billing entity, with discounts starting at 6, 11, and 26 seats. Reassigning a seat to a different user or device is simple: open the license management page and activate the license on the other device.

What happens if I cancel my subscription?

Your license remains active until the end of the period you have already paid for. After that, the app no longer allows new anonymizations, but you keep access to the documents you have already produced. You can reactivate your subscription at any time and resume using Celatum on the same device. No separate data-export step is needed: all documents and generated data are already stored locally on your computer.

Which languages and personal identifier types does Celatum recognise?

The detection model is trained on multilingual text and works on documents in German, French, Italian, and English (the four Swiss working languages). It recognises 50+ identifier types including persons, organisations, addresses, AVS numbers, IBANs, dates of birth, phone numbers, email addresses, court references, case numbers, and more.

Download Celatum

Audited release · v1.0.1

This exact build passed an independent security review. Security statement (PDF)

Intel Mac users: contact us for an x86_64 build.

System Requirements

  • macOS 12 (Monterey) or later
  • Windows 10 or 11 (64-bit)
  • 6 GB RAM minimum
  • 1 GB disk space for model files

Disclaimer: This application is designed to assist with personal identifier detection and anonymization, not to replace human judgment. All output should be reviewed by a person before use. The software may produce false positives or miss certain data. The user bears full responsibility for verifying the accuracy and completeness of documents processed with the software.