Mask personal data in text, documents and tables before it reaches AI, analytics and testing systems.
October 2026 · Employee pay statement
HR-2026/118 · 17/10/2026| Pay item | Amount |
|---|---|
| Gross pay | GBP 4,500.00 |
| Deductions | GBP 1,020.00 |
| Net pay | GBP 3,480.00 |
Pay items stay useful; employee identity and bank details are masked.
Fictional document · No real personal or company data
Support data minimization by masking personal data before sharing, analytics and AI workflows.
Raw files and restore mappings are not written to processing history; users download the results directly.
New PDF jobs retain encrypted original and redacted copies for authenticated history access. On-premises storage and retention are agreed for your deployment.
The SaaS setup can be positioned for EU-region hosting, while the on-prem option remains available.
Review transport, storage, administrator access and retention before choosing a deployment. Our Privacy Policy explains the current SaaS safeguards and limitations; enterprise controls are agreed separately.
Browser and API access use HTTPS. Archive copies are encrypted before transfer; the archive FTP channel itself does not currently use TLS. See the Privacy Policy for details.
Original and redacted PDF archive copies are encrypted before storage. On-premises storage requirements are scoped with your infrastructure team.
Encryption keys and document data are managed separately. Restoration mappings are used only in workflows that need them, with limited access.
Depending on the deployment model, recognition services can run locally or in isolated environments without requiring documents to be sent to public LLM providers.
Upload, scan, mask and download steps run automatically. Operational access is minimized and limited by the organization’s policy.
SaaS PDF archives currently have no automatic expiry. Contact us for archive deletion. Enterprise retention requirements are agreed as part of deployment planning.
With SaaS, private cloud or on-premises deployment options, data location, network access and capacity model can be selected for your needs.
The searched entity types, applied methods and output timing can be tracked without unnecessarily logging sensitive content.
Replace personal data in call notes, emails and complaints with consistent tags. Restore tagged AI responses using the mapping.
Read scanned pages. Redact personal fields, faces, signatures and QR codes, then download a masked PDF.
Discover personal data columns automatically. Use format-preserving synthetic values, tags or partial masking.
Use Z-Mask for a single text or a recurring data pipeline. The web workspace and API use the same masking services.
Sign in to process text, PDFs, images or tables. Choose masking settings, download the output and review your processing history.
Open the workspaceMask multiple texts in a JSON array with one request. Process many records in CSV, Excel and SQLite files using column-specific methods.
Try batch processingSend requests directly from CRM, document management, analytics and AI workflows. For document collections, build a workflow that submits each file in sequence.
Discuss API integrationCRM notes → batch text masking · Document archive → per-file redaction · Test data → column masking · LLM → mask, then restore the response with a mapping
From one-off document masking to team workflows: try it with free page credits, then add more page credits through Stripe when needed.
For personal trials and small one-off tasks.
For small teams with recurring document redaction needs.
For companies processing high-volume document archives for compliance.
Enterprise deployment on your own infrastructure, with installation and integration scoped to your organisation.
From banking to healthcare and legal files to customer conversations: a masking workflow that keeps data useful.
Payroll review for Emma Clarke shows net pay of GBP 3,420.18 sent to sort code 20-45-78 and account number 11876543.
The same file includes National Insurance number QQ 12 34 56 C and a complaint email from emma.clarke@example.com about a delayed refund reference.
Request: share the payment timeline with finance and legal while keeping payroll identifiers out of analytics and AI summaries.
Salary amount, complaint context and payment status remain usable; names, sort codes, account numbers and NI references are masked.
Patient Oliver Bennett called NHS 111 after chest tightness; the record lists NHS number 485 777 3456 and date of birth 14/02/1979.
Dr. Hannah Price advised urgent assessment if symptoms return and asked that discharge notes are not emailed to oliver.bennett@example.com.
Callback number: +44 7700 900123. Clinical risk, medication class and follow-up interval can remain visible for review.
Symptoms, risk level and follow-up plan remain readable; patient identifiers and contact data are redacted before sharing.
The requester James Whitfield asked for interview notes and internal correspondence linked to case HR-DSAR-4491.
Panel comments mention line manager Rachel Morgan and include a handwritten note from Tom Hughes about performance concerns.
Before disclosure, the reviewer must decide which third-party names, email addresses and opinion notes should be redacted under UK GDPR.
The factual timeline stays intact; staff names, third-party identifiers, addresses and private notes can be consistently hidden.
Sophie Martin: Please delete the old support tickets on account AC-448291; they include my home address 42 Camden Road, London NW1 9XY.
Agent: I confirmed contact details as +44 7700 900456 and sophie.martin@example.com before escalating the erasure request.
Outcome: retention basis and deletion status can stay visible for audit, while personal fields are masked in exports and analytics.
Ticket category, SLA and resolution remain clear; names, address, email, phone and account identifiers are removed.
Tag personal fields before calling a model for call summaries, email replies or document analysis. Restore response tags using the mapping data.
Your application receives a call note, email or customer request.
Selected fields become consistent tags such as [PERSON_1] and [PHONE_1].
A cloud or on-premises model generates a summary or reply from the masked content.
Use the mapping in your own workflow to restore tags to their original values.
Your organization manages access to mappings for reversible tagging. For test data, choose other methods such as synthetic values or hashes.
Read text with OCR and mask faces, signatures, QR codes and barcodes alongside personal fields. Review the boxes, add missing areas manually and download the result.
Names, ID numbers, dates of birth, photos and machine-readable passport fields.
Rights holders, family details, identity fields and signatures in document-specific layouts.
Personal fields in dense text, multi-row tables and rotated scans.
Headers, engineer names, phones, addresses and signatures while preserving the drawing's technical content.
A validation rule engine works alongside a multilingual recognition model. Personal details are masked while organization and place names remain readable.
Define your own data types without retraining or code changes.
medical diagnosis → [DIAGNOSIS_1]CUST-482913 → [CUSTOMER_ID_1]Project Atlas → [PROJECT_1]Profile columns in CSV, Excel and SQLite files. Choose a method for each data type and use case; preserve relationships through consistent transformations.
| Method | Example output | Use case |
|---|---|---|
| Synthetic value | Emma Clarke | Realistic test data with valid formats for supported types. |
| Tag | [PERSON_1] | Track people and record relationships with readable tags. |
| Partial masking | E*** C***** | Keep only the necessary part of a value visible. |
| Hash | 7f42a… | Create a consistent representation without revealing the original value. |
| Redaction | ████████ | Hide the field's contents entirely. |
Connect the workspace's masking engine to your own applications. Use a shared service layer for text, document and table processing.
Multilingual recognition with validation rules for EU national IDs, IBANs and card numbers.
New entity types through descriptions, regular expressions and term lists, with selectable types and thresholds.
Start with CPU, accelerate with GPU and process in parallel as needed. Measure capacity with your own data during a pilot.
We install Z-Mask on your organisation's own servers or private cloud. Contact our team to discuss your infrastructure, document volume and integration requirements.
Offline masking, local models and your own policies. Start with CPU and add GPU capacity as needed.
Run in your own cloud account. Scale capacity horizontally with stateless services.
Sign in, process your text or documents and manage results in one workspace.
Discuss deployment scope, hardware requirements, integration and support with our team. Enterprise installations are quoted individually, separately from online subscriptions.
Define success criteria together: which details to hide, which context to preserve and how to handle your workload.
Measure recognition quality and speed with sample data; define entity types and masking policies.
Connect the API to authentication, LLM, document and database workflows.
Finalize deployment, capacity and monitoring, then bring real workloads online.
Improve policies with new data types and tags; review quality periodically.
Timelines are indicative and will be agreed based on your scope and integration needs.
Let's discuss a pilotWhen text tagging and mapping are enabled, use the mapping to restore tags. Synthetic values, hashes and redaction do not use this restoration workflow.
Yes. Use a regular expression for customer numbers, a term list for project names or a description-based entity type for diagnoses.
Compare original and masked pages, remove masking boxes or add new ones manually, then regenerate the PDF.
The workspace stores account-linked summaries, masked text and encrypted copies for newly archived PDF jobs. Account owners and authorized administrators can access records. Deleting a history entry does not necessarily delete archived files; contact us for archive deletion. Older jobs and original image inputs may lack copies.
On-premises deployment with local models is supported. Hardware, capacity and integration requirements are evaluated during a pilot.
Start a pilot with your own data.