Typly Anonymizer — document anonymization for law firms (online and offline)

How it works

Three steps to a finished document

A similar flow online and on the computer. You decide what gets covered.

Load

Paste text or load a file — PDF, an Office document, a scan or a photo. The app detects personal data: PESEL, NIP, names, addresses, case references and more.

Review and fix the highlights

You review the highlights, remove false positives and add anything missed. Reviewing the result is a natural step — you approve what is finally hidden.

Save or prepare for AI

Save the document with permanent redaction — or swap the data for reversible [PERSON_1] tokens to work with ChatGPT, Claude or Gemini and restore them later. The reversible mapping kept locally is available in the desktop edition.

Two editions

Online or local — pick how you work

Same goal: detect personal data and prepare the document. What differs is where it is processed and how you get access. A dash means we don't claim that feature for this edition in this table.

Feature	Online — anon.typly.app	Desktop — Windows / macOS
How you start	In the browser, sign in with Google or Microsoft	Install the app on your computer — access on request
Where data is processed	On our server	Fully local, offline — nothing leaves the computer
Input	Text and files uploaded in the browser	PDF, DOCX, PPTX, ODT, RTF, scans and photos (JPG / PNG / TIFF)
Output	Document prepared to download / use with AI	PDF — permanent redaction or reversible tokens
Preserves original layout	Yes — keeps the document layout	Office files are rendered to A4 pages (output is PDF)
OCR for scans and photos	Yes	Yes — offline, included
AI mode (reversible tokens)	Yes — mapping on the server	Yes — [PERSON_1], mapping locally on your computer
Best for	A quick, self-service test	Day-to-day work with confidential firm documents

Try it online → Request an offline test →

Online: the free plan covers 5 documents. Desktop: a 30-day trial on request (for law firms, a pilot of up to 90 days is possible once scope is agreed).

The problem

Your organization uses ChatGPT or Claude. Do you know what happens to personal data?

In most Polish companies and law firms — nobody does.

Employees use AI off the books

A lawyer pastes a contract into ChatGPT to speed up drafting. HR asks Claude to evaluate a CV. Accounting summarizes an invoice via Gemini. Polish PESELs, NIPs and client names end up in external services. The person responsible for data protection is the last to find out.

Manual redaction costs hours

Preparing a single letter for sharing can take 15+ minutes. Hundreds of letters per month add up to real effort. A law firm pays lawyer rates for work that can be sped up.

Accountability under GDPR and the AI Act

Using AI without control over data is a risk. GDPR, the AI Act and sector rules expect accountability. Existing processes rarely account for AI scenarios.

Use cases

What law firms and offices use it for

One product, several use cases — from working with AI to sharing documents.

Safe work with AI

You swap personal data for reversible tokens, send the document to ChatGPT, Claude, Gemini or Grok, and restore the original values once you get the answer.

In the desktop edition the token mapping stays locally on your computer.

A document ready to share

Preparing letters for external sharing or publication. We detect Polish data — PESEL, NIP, IBAN, names, addresses, case references — and permanently redact it.

PDF, Office or scan in → PDF out. Office files are rendered to A4 pages — we do not reproduce the original editable layout.

A repeatable, controlled process

Instead of manual blacking-out — a repeatable flow: detect, review the highlights, save. You always review and approve the highlights yourself before saving the document.

Capabilities

What sets Typly Anonymizer apart

Polish PII — recognized in context

14 categories of Polish identifiers: PESEL, NIP, REGON, national ID, IBAN, KRS, administrative case number, land register number, court case reference, postal code, phone, date, e-mail, company names with legal form.

Numbers are verified with dedicated rules — not just "some 11-digit string", but a PESEL after a checksum. Names, addresses and companies are recognized in Polish linguistic context. No detection is perfect — which is why you always review the result before saving.

PESELNIPREGONNational IDIBANKRSCase numberLand registerCourt case refPostal codePhoneDateE-mailCompany name

You load (desktop)	You get
PDF (text)	PDF — redaction or tokens
DOCX, PPTX, ODT, RTF	PDF (rendered A4 pages)
Scans and photos (JPG / PNG / TIFF)	PDF after offline OCR

Office files are rendered to A4 pages and saved as PDF — we do not reproduce the original layout, tables, fonts or comments of the editable file.

Formats: what you load and what you get

In the desktop edition you can load text files, Office documents, and scans and photos (with offline OCR). The output is always a PDF — with permanent redaction or reversible tokens.

Two output modes + engine strategies

The desktop edition offers two working modes:

Permanent redaction — data removed with no reverse path. For sharing.
Reversible tokens — stable [PERSON_1] with a local mapping. For working with AI and restoring the data.

On-premise and API deployments add further engine strategies (including a deterministic salted hash and format-preserving numbers) — described below, in the section for teams.

redaction

Permanent redaction

Data removed, no reverse path. For sharing. (desktop)

[PERSON_1]

Reversible tokens

Stable token with a local mapping. For working with AI. (desktop)

hash

Hash

Deterministic pseudonym from a salt. For logs and consistency. (engine / on-premise)

keep_format

Keep format

Number structure preserved, value gone. (engine / on-premise)

Online edition

Try it in the browser — no install

The fastest way to check it yourself on your own document. Processing happens on our server.

Free plan · 5 documents free

Sign in with Google or Microsoft and try the product on your own documents — no card, no commitment.

Open anon.typly.app →

In the online edition, files are processed and stored on the server in encrypted form. We never use them to train models.

Request an offline test

Test Typly Anonymizer locally at your firm

The desktop edition runs entirely on your computer. After your request we'll provide access to the Windows or macOS installer plus hands-on help at the start. Standard trial is 30 days; for law firms a pilot of up to 90 days is possible once scope is agreed.

You'll learn the terms and price for continued use before the pilot begins.

The data you provide (email address, organization name, number of seats, system and message) is processed solely to handle your request for an offline test. Data controller: Typly Sp. z o.o. Details and your rights: privacy@typly.app.

Mobile app · free

Anonymize right on your phone — fully offline

A separate, free mobile app. Black out photos, PDFs and scans, or swap data for tokens before pasting into AI — all processing happens on the device.

100% offline — nothing leaves the phone, works in airplane mode
Permanent burn-in redaction + token mode for safe AI use
No account, no login, no ads or trackers

The mobile version is a lightweight, on-device offline engine. The full engine (14 categories, OCR, wider format range) is in the online, desktop and on-premise editions.

Privacy

Where your documents go

We separate it clearly: the offline edition stays with you, the online edition processes and stores files on the server.

Desktop edition (offline)

All anonymization happens on your computer. Documents, detected data and the reversible-token mapping never leave the device — the app works without an internet connection.

Because we don't receive your documents, in this edition we are not a processor of your data.

Online edition (anon.typly.app)

This is a web app — the files you upload are processed and stored on the server so the repository and document work can function. We keep the data encrypted and never use it to train models.

If you work with especially sensitive documents, choose the desktop edition or an on-premise deployment.

Metadata and your rights

To run accounts and limits we keep minimal technical metadata (e.g. usage counters, timestamps). We do not publish or sell your data.

GDPR rights — access, deletion, portability, objection — are handled by writing to privacy@typly.app.

Data controller: Typly Sp. z o.o. · Data contact: privacy@typly.app · Right to complain: the Polish DPA (UODO).

For teams

On-premise and API — deployment for your team

On-premise

Inside your network. Full control.

Full product inside your infrastructure
Your documents never leave your network
Team accounts, repository and audit log
Configuration aligned with your security policy
Hands-on support from our team during deployment

Recommended for: public sector, banks, insurance, healthcare, regulated industries.

Hosted / API

No infrastructure to maintain.

REST API and Web UI from your account
Processed on our server in the European Union (EEA)
Extra engine strategies (hash, keep_format) and audit trail
Scaling managed by us
Updates and new features rolled out automatically

Recommended for: teams without dedicated DevOps, pipeline integrations, API scenarios.

Let's talk about deployment →

Industries

Safe AI for different industries

⚖️

Law firms

Masking client data in letters before sending to ChatGPT or Claude. Detection of court case references, KRS and land register numbers. Preparing PDF documents for sharing with permanent redaction.

🏛️

Public administration

Redaction before sharing and publication. Responses to public information requests. Clerks using AI assistants without sending resident data outside the organization.

🏢

Compliance and IT teams

Bringing order to AI use across the organization. Employees use AI officially and safely. Polish PESELs don't leak uncontrolled to external services. On-premise or pipeline deployment.

🏥

Regulated sectors

Banks, insurance, healthcare. High compliance bar plus pressure to adopt AI. On-premise deployment, a data processing agreement and configuration tailored to the organization's processes.

Engine lineage

Engine from the INFOSTRATEG IV (NCBR) programme

We developed the Typly anonymization engine within the INFOSTRATEG IV research programme run by NCBR (the Polish National Centre for Research and Development) — a strategic programme advancing modern technologies for Polish public administration.

In the programme, more than a million documents from Polish public administration were processed: resident inquiries, invoices, administrative decisions and official correspondence. A realistic, varied corpus — not a synthetic test set.

The desktop edition shares the same lineage — Polish rules, dictionaries and identifier validation. It was not separately tested on that corpus, so we encourage you to check the detection scope and quality on your own documents during the offline test. We do not disclose partner office names, in line with our agreements.

1,000,000+

documents processed in the INFOSTRATEG IV programme

Compliance

Support for GDPR and the AI Act

The tool supports compliance processes — it does not replace your organization's legal assessment. Some of the functions below apply to the engine and to on-premise and API deployments.

Pseudonymization in the spirit of GDPR Art. 4(5)

Reversible tokens with a mapping match the idea of pseudonymization (mapping is local in the desktop edition).

Permanent anonymization for sharing

Permanent redaction removes data with no reverse path — for letters shared externally.

Audit trail (engine / on-premise)

In API deployments each operation can return a list of detected entities with positions — material for accountability.

Support for the AI Act

Minimizing personal data before passing it to AI is part of good practice for high-risk systems.

Reducing leaks via metadata

Document metadata and file names containing personal data are neutralized.

Configurable salt (on-premise / API)

The hash strategy uses your salt, so pseudonyms don't collide between organizations.

See Typly Anonymizer on your documents

A 15-minute live demo with our technical team. We'll show the product on a document from your organization (anonymized in advance for the test). No sales slides.

Pick a slot ↓ Pobierz folder (PDF)

The demo is run by our technical team. We sign an NDA before any demo on your documents. Standard DPA and contract templates available on request.

Access and terms

How to get started

The mobile app and the desktop test are free. You'll learn the terms and price for continued use before the pilot begins.

📱

Mobile app

Free

iOS and Android in the stores — no licence, no account.

🌐

Online (anon.typly.app)

Free plan to start

5 documents free. Sign in with Google or Microsoft.

💻

Desktop (Windows / Mac)

30-day test

Runs fully locally. Access on request via the form.

🤝

Law firm pilot

Up to 90 days

Once scope is agreed: agreed seats, help at the start, gathering feedback.

You'll learn the terms and price for continued use (desktop, on-premise, API) before the pilot begins. Additional data categories and rollout are agreed in the conversation.

FAQ

Frequently asked questions

How does the online edition differ from desktop?

Online (anon.typly.app) runs in the browser and documents are processed on our server — the fastest way to test it yourself. Desktop runs fully locally on your computer (Windows / macOS), offline, and we provide installer access after your request.

Does the desktop edition preserve the layout of a Word or PowerPoint file?

No. Importing Office files (DOCX, PPTX, ODT, RTF) means extracting the text and rendering pages to A4 — the output is a PDF. We do not reproduce the original layout, tables, fonts or comments of the editable file.

What about scanned documents or photos?

The desktop edition has built-in offline OCR (included), so it also anonymizes scans and photos (JPG, PNG, TIFF). You get the result as a PDF.

Are my documents stored?

In the desktop edition — no; everything stays on your computer and we don't receive your files. In the online edition, documents are processed and stored on our server (in encrypted form) so the repository works — we never use them to train models.

Do you detect all personal data?

No automatic detection is perfect. That's why you always review and fix the highlights before saving — you approve the final result. The detection scope is best checked on your own documents during the test.

Can I test without talking to sales?

Yes. You can launch the online edition right away at [anon.typly.app](https://anon.typly.app) — sign in with Google or Microsoft, 5 documents free. You request the offline test via the form on this page.

How long is the test and what does it cost?

Desktop is provided for 30 days as standard. For law firms a pilot of up to 90 days is possible once scope is agreed. You'll learn the terms and price for continued use before the pilot begins.

Do I have to sign a Data Processing Agreement (DPA)?

It depends on the edition. Desktop and on-premise run fully locally — your data never reaches us, so we are not a processor and no DPA is required. In the online edition (anon.typly.app), where documents are processed on our server, we settle the processing arrangement before processing begins.

Company & IP

Polish company · IP registered in the US and EU

For procurement and compliance — full due-diligence:

Company

Typly Sp. z o.o.

HQ: al. Gen. W. Sikorskiego 9B u3D, 02-758 Warsaw, Poland

Trademark

USPTO Reg. No. 7027595

TYPLY®, granted 2023-04-11

EU Community Design

EUIPO RCD No. 009171887

Keyboard interface, valid through 2027-09-14

R&D project

INFOSTRATEG IV (NCBR)

No. INFOSTRATEG4/0011/2022, PLN 3.28M funding

Contact

Let's talk

Live demo

Pick a slot

Business contact

hello@typly.app

Data Protection Officer

privacy@typly.app

Prepare your firm's documents for sharing and for working with AI