The most effective way to protect personal data from an AI provider is not to send it in the first place.
Why we built it
Generative AI tools have become everyday instruments in law firms, finance departments and HR teams: drafting, summarising, translating and reviewing documents. The same documents, however, are full of names, addresses, tax numbers, bank account numbers, health insurance identifiers and project code names. Pasting them into a chat window means transferring personal data — and often trade secrets — to a third-party controller, frequently outside the European Union, with little visibility into how the data is stored, used for training or disclosed.
As a boutique firm advising on data protection and AI regulation, we faced this tension in our own practice. We wanted to use AI tools responsibly, without exposing client data. Maskovski (maskovski.com) is the result: a practical legal-tech tool that lets the benefits of AI be used while the personal data stays where it belongs — with you.
What it does
Maskovski runs entirely in your browser. You open a Word (.docx), PDF or text file, or paste text, and the tool detects and replaces identifying elements with consistent placeholders such as [PERSON_1], [COMPANY_1], [ADDRESS_1] or [ACCOUNT_1]. The anonymised text can then be copied into any AI service or saved as a new Word file. A key table stored only in your browser tab lets you restore the original names in the AI's answer with one click.
- Rule-based detection tuned to Hungarian and English names, company forms and identifiers: personal names, company names, addresses and land-registry numbers, e-mail addresses, telephone numbers, IP addresses, tax numbers, bank account numbers, social-security and ID-document numbers, licence plates, dates of birth and, optionally, all dates and monetary amounts.
- Custom terms: project and product names, technical parameters and other business secrets can be added manually or by selecting text in the source document.
- Full control: every detected element can be unticked (for example your own company name), and the key table can be exported to a file for later re-identification.
- Word files keep their formatting, tables, comments and tracked changes, while hidden metadata — author, last saved by, company, thumbnail — is removed and the file receives a neutral name.
- Local processing: files are never uploaded to a server. The tool has no account, no login and no backend database.
Pseudonymisation, not anonymisation — and why that matters
We deliberately describe what Maskovski does as pseudonymisation. As long as the key table exists on your side, the data remains personal data in your hands and the GDPR continues to apply to you as controller. What changes is the position of the AI provider: it receives text without the key. Under the Court of Justice's 2025 judgment in EDPS v SRB (C-413/23 P), such data is not personal data for the recipient where the recipient cannot reasonably re-identify the data subjects. Maskovski is designed around that test — minimising what the provider receives, and keeping the means of re-identification exclusively with the user.
This is also how we understand the principles of data minimisation and data protection by design in Article 5 and Article 25 GDPR: the technical measure is applied at the point where the risk arises, before the data leaves the organisation, rather than relying on contractual promises after the fact.
The limits we want users to know about
Responsible use also means being candid about what a tool cannot do. Automated detection is rule-based: rare names, abbreviations and typos may be missed, and unique facts — a job title, a diagnosis, a small town, a floor area — can identify a person even without a name. Maskovski does not recognise such contextual identifiers, nor the content of embedded images. For that reason the tool asks users to read through the output before uploading it anywhere, and it reminds them never to paste the key table into an AI. Maskovski is a technical aid; it does not replace a legal assessment of whether a given use of AI is lawful in a given context.
Our commitment
At DPZ Legal we advise clients on the GDPR, the AI Act and the governance of AI in organisations. Maskovski reflects the same principles we recommend to our clients: use AI, but keep personal data and trade secrets under your own control; prefer technical safeguards over paper ones; and be transparent about residual risk. We are making the tool available free of charge because we believe responsible AI use should be the default, not a premium feature.
Try it at maskovski.com. If you would like help designing an AI-use policy, a data-protection impact assessment for an AI deployment, or an AI Act readiness review, we would be glad to talk.
Disclaimer: Maskovski is provided as-is and does not constitute legal advice. The user remains responsible for verifying the output before sharing it.