4 practical scenarios with before/after examples
DeepL for translations, Copilot for email summaries, Claude for report analysis, ChatGPT for proposal templates — AI tools have become indispensable in everyday work. Yet every time a document containing personal data is submitted to one of these services, a GDPR risk arises.
This article demonstrates with four concrete practical scenarios how pseudonymization solves the problem — with before/after examples for each scenario.
Translate international contracts without data protection risk
The problem: A German-language purchase contract needs to be translated into English for an international partner. The contract contains names of the contracting parties, addresses, account numbers and contract amounts.
"Between Mueller Consulting Ltd, represented by Dr Thomas Mueller, 42 Station Road, London EC1A 1BB, and Schmidt & Partners KG, 7 High Street, Vienna 1010, the following contract is concluded..."
All personal data is transferred to DeepL servers.
"Between Company_A, represented by Person_A, Address_A, and Company_B, Address_B, the following contract is concluded..."
DeepL translates the pseudonymized text. The pseudonyms are preserved as placeholders. After translation, they are replaced with the original data in the translated text. Result: perfect translation with correct names and addresses.
Efficiently distil lengthy email threads to their essentials
The problem: A lengthy email thread with 30+ messages needs to be summarised for a management meeting. The emails contain sender and recipient names, email addresses, phone numbers in signatures and confidential project details.
"From: sarah.weber@company.co.uk
To: michael.brown@partner.com
Subject: Project Alpha — Milestone 3
Dear Mr Brown, please find attached the updated project plan. Please forward to Dr Klein (Tel: +44 7700 900123)..."
"From: Email_A
To: Email_B
Subject: Project Alpha — Milestone 3
Dear Person_B, please find attached the updated project plan. Please forward to Person_C (Tel: Phone_A)..."
Copilot summarises the pseudonymized email thread. The substantive key points are preserved while no personal data is transferred. After summarisation, the pseudonyms are re-identified.
Have medical and legal reports securely reviewed by AI
The problem: A medical report needs to be checked by Claude for completeness and plausibility. The report contains patient names, dates of birth, diagnoses, treatment histories and names of attending doctors.
"Patient: Maria Hofmann, DOB 15/03/1978
Diagnosis: Herniated disc L4/L5
Treatment: Dr Andreas Wagner, Orthopaedics Munich
The patient first presented on 22/01/2026..."
"Patient: Person_A, DOB Date_A
Diagnosis: Herniated disc L4/L5
Treatment: Person_B, Orthopaedics Location_A
The patient first presented on Date_B..."
Claude can fully analyse the report — the medical content is preserved, only the personal data is pseudonymized. The AI checks completeness, plausibility and formal correctness without knowing the identity of the individuals involved.
Generate personalised proposals based on existing templates
The problem: A sales representative wants to use ChatGPT to create a personalised proposal based on an existing template. The previous proposal contains client names, company address, specific prices and terms.
"Proposal No. 2026-0847 for Maier Logistics plc
Contact: Ms Lisa Gruber, l.gruber@maier-logistics.com
Licence fee: EUR 24,500/year
Discount: 15% (special terms agreed with Mr Maier)..."
"Proposal No. Reference_A for Company_A
Contact: Person_A, Email_A
Licence fee: Amount_A/year
Discount: 15% (special terms agreed with Person_B)..."
ChatGPT can analyse the proposal structure and suggest improvements without knowing the real client data or prices. After re-identification, the sales representative has an optimised proposal with all original data.
Docuflair Mask automatically pseudonymizes your documents before they are sent to DeepL, Copilot, Claude or ChatGPT. Experience the workflow in a 15-minute demo.
Answers to the most important questions about sending documents securely to AI tools
Yes, DeepL Pro offers enhanced data protection: texts are deleted after translation and are not used for training. However, data is still transferred to DeepL servers. With pseudonymization, you can use even the free version safely as no real personal data is transmitted.
With Microsoft 365 Copilot on the Enterprise plan, data is processed within the Microsoft tenant. Nevertheless, pseudonymization is recommended for particularly sensitive contracts as data flows through Microsoft systems and is subject to US law.
Health data belongs to special categories of personal data under Art. 9 GDPR and enjoys particular protection. Pseudonymization is especially important here: all patient names, dates of birth, diagnoses and treatment data are replaced with pseudonyms before the document is submitted to the AI.
Yes. The pseudonyms (Person_A, Address_A etc.) are recognised by translation AIs as placeholders and are not translated. After translation, the pseudonyms in the translated text are replaced with the original data. The result is a complete translation with correct original names.