How to use PrivacyScript
PrivacyScript finds names and other personal details in health documents, interview transcripts and survey data, and removes or replaces them. Everything runs in your browser. Your files are never uploaded.
Anonymise or pseudonymise?
Removes identities for good. Nothing can be reversed. Best for AI tools and anything shared outside your organisation. Output is blocked if people could still be singled out.
Replaces identities with codes. You get a key file, locked with your passphrase, to reverse them later. The data is still personal data under GDPR, so keep it inside your approved research or care setting.
Check a document
Find out what is in a document before you share it or paste it into an AI tool.
- 1On the home page choose "Check".
- 2Pick the rules that apply to you (EU, UK, US or General).
- 3Drop your file. A progress card shows each stage while it is scanned.
- 4Read the report. If it is not safe, choose Anonymise or Pseudonymise to clean it up.
Clean up a document
- 1On the home page choose "Clean up" and pick where the data will go.
- 2Drop your file and wait for the scan (the first time, the name model downloads once, about a minute).
- 3Work through the numbered review steps. Only the current step is open; press its button to move on.
- 4Read the risk check. If the risk is high you must confirm before continuing.
- 5Download the cleaned file and the audit log. For pseudonymise, also save the key file with a passphrase.
Interview transcripts
Works with TXT, Word, VTT and SRT, including Zoom and Teams exports, in English, Dutch, German and Spanish. Timestamps are kept.
- 1Check the speakers: real names become "Interviewer" or "Participant 1". Change a role with one click, or type your own label.
- 2Choose readable labels ([Person 1]) or codes. Readable labels keep the same number for the same person.
- 3Decide on each highlighted passage. Some sentences identify people without a name, such as "the only nurse on the ward". Keep or remove each one.
- 4Decide on possible names we were not sure about. Confirming a name replaces every mention of it.
Surveys and spreadsheets
Works with CSV, Excel and SPSS files, including Qualtrics, REDCap, SurveyMonkey and Microsoft Forms exports. The result downloads as CSV.
- 1Check what each column contains. Columns that identify people (names, emails, IP addresses) are removed or coded.
- 2Look at "Can anyone be singled out?". Each person must share their details with at least 5 others.
- 3If not, press "Fix automatically". It widens groups, for example exact ages become age bands.
- 4Confirm the columns and continue.
Many files at once
Use Batch processing for a study with several interviews. The same person gets the same label in every file. You review all flagged passages on one screen, and files that fail a check are held back. Open surveys and PDFs one at a time.
Reversing pseudonymised data
Go to Open a key file, choose your key file and enter its passphrase. You can look up any code, see how dates were shifted, and restore a pseudonymised text file. Keep the key file separate from the data; without the passphrase it cannot be opened, and it cannot be recovered if lost.
Before you share
- Automated checks can miss things. Skim the result, and ask: could someone who knows this person recognise them?
- Check that your ethics approval and data management plan allow sharing, especially with AI tools.
- Keep the audit log with your records. It shows what was changed and contains no personal data.
- Text in scripts PrivacyScript cannot read (for example Tamil or Sinhala) cannot be checked. It will tell you.
Frequently asked questions
Is my data uploaded anywhere?
No. PrivacyScript runs entirely in your browser. Files are read and processed on your device and never sent to a server. Only the name-detection model is downloaded once, and it is then stored by your browser.
What is the difference between anonymising and pseudonymising?
Anonymising removes identities for good and cannot be reversed, so the output can be used with AI tools or shared publicly. Pseudonymising replaces identities with codes that you can reverse with an encrypted key file; the data is still personal data under GDPR.
Can I de-identify interview transcripts before using ChatGPT or other AI tools?
Yes. PrivacyScript reads TXT, Word, VTT and SRT transcripts, including Zoom and Teams exports. Speaker names become Interviewer or Participant labels everywhere, other names become [Person 1]-style labels, and sentences that could identify someone by context are flagged for you to keep or remove.
Which survey formats are supported?
CSV, Excel (.xlsx) and SPSS (.sav) files, including Qualtrics, REDCap, SurveyMonkey and Microsoft Forms exports. PrivacyScript measures whether any respondent can be singled out (k-anonymity) and can widen groups automatically, for example exact ages to age bands.
Which languages does it support?
Name detection works in English, Dutch, German, Spanish, French, Italian and Portuguese. Transcript rules for context and family names cover English, Dutch, German and Spanish.
Is it compliant with GDPR and HIPAA?
PrivacyScript supports GDPR pseudonymisation (Article 4(5)) and anonymisation (Recital 26), HIPAA Safe Harbor, UK GDPR and the European Health Data Space. It produces an audit log for every run. Automated checks can miss context, so review the output before sharing and follow your ethics approval.
Is PrivacyScript free?
Yes. It is free to use in any modern browser, with no account and no tracking.