Skip to content
Product6 min read

UK Legal Documents in ONYRI Sanitize: What Gets Masked, and What You Still Check

ONYRI Sanitize masks names, addresses, case numbers, NHS and National Insurance numbers in UK legal papers. See each format, and the gaps you still check.

By Alexis de ONYRI

ONYRI Sanitize masks names, addresses, case numbers, Home Office references, NHS numbers, National Insurance numbers and bank details in a UK legal document. It does this only once it recognises the document as British, and it misses some things. It never replaces the review by the fee earner, the solicitor or caseworker who handles the file.

We checked this in November 2026 by running its detection rules, without the name-recognition model, on fictitious documents. Every person, number and case below is invented.

How does ONYRI Sanitize know that a document is British?

The Default profile reads the whole document and picks a country. These clues count: a £ amount, a UK postcode or a +44 number. So do words such as United Kingdom, HMRC, National Insurance, NHS, sort code or Ltd. A few court names count, in their usual spelling, such as “Employment Tribunal”, “First-tier Tribunal” or “Crown Court”.

A heading in capitals does not count: “IN THE EMPLOYMENT TRIBUNAL” is no clue, nor are “Home Office” alone or “Solicitors”. With no clue, the site's language decides: the English version uses US rules, or UK rules in a browser set to English (United Kingdom), and the French and German versions use French or German rules. On Pro, a custom profile keeps the country you choose.

Which UK formats are masked, and which need their label?

Some formats are masked on sight. Others look like ordinary numbers, so the tool waits for a label in front, such as “Passport number:”.

FormatFictitious exampleWhen it is masked
UK Visas and Immigration case referenceGWF071234567On sight
Immigration appeal numberHU/12345/2025On sight
Employment Tribunal case number1401234/2026On sight
NHS number123 456 7881On sight, if the last digit checks out
National Insurance numberJG 10 20 30 DOn sight
Sort code and account number together00-00-00 12345678On sight
PassportPassport number: 543219876After its label
Biometric residence permitBRP number: ZR1234567After its label
Home Office referenceHome Office reference: A1234567After its label
NHS number with a wrong last digit123 456 7882Not detected
Unique Taxpayer Reference, short labelUTR: 1234567890Not detected
Tested with the country read as United Kingdom.

A nine-digit passport number with no label stays visible. Amounts in £ and dates are masked by default, so untick any value the recipient needs in “Detected data”. HMRC’s National Insurance manual says D, F, I, Q, U and V are never used in the first two letters. So “QQ 12 34 56 C” is left alone.

How does it find names in a Home Office letter or a witness statement?

Most names are found by how they are written. On a computer, a name-recognition model that runs on your device finds more. It does not run on a phone or a tablet.

  • Party names in capitals in a court heading, such as “MS JANE EXAMPLE” then “Claimant”.
  • Names after a judicial title or a label: “Judge Smith”, “Claimant: Jane Example”.
  • Names in a set phrase, such as “my colleague, David Sample”.
  • A full name found once is masked wherever it returns, in capitals too. In a UK document each part is also masked alone, except common words.

Street addresses with a number are masked, and so is every postcode. The other lines of an address are masked in some layouts only: a town alone on its line can leave the lines above it visible. A court’s or public body’s address is mostly left alone. Its postcode is always masked, though, and its numbered street line can be masked too. In our Home Office letter, the office’s street line and postcode were masked, not the building name. Check the preview.

How does Token mode label people across a bundle?

In Token mode, each value becomes a label such as [NAME1]. A person usually keeps one label on every page, even when the name is shortened. Token mode comes with Pro and the 7-day pass, not with the Free plan or without an account. There is no “restore the original” button and no key file, so keep your original.

In the UK GDPR, point 5 of Article 4(1) defines pseudonymisation. It means processing after which data cannot be attributed to a person without extra information kept separately. Token mode is pseudonymisation, not anonymisation: treat the text as personal data.

  • With only one Jane Example in the file, “MS JANE EXAMPLE”, “Ms Example” and “Jane” share one label.
  • If “Karim Example” also appears, a bare “Ms Example” could mean either person, so it gets its own label.
  • A titled surname such as “Mr Sample” joins Peter Sample only if “Mr Peter Sample” also appears.
  • A possessive such as “Ms Example’s” gets a label of its own.

The rule is cautious on purpose: two labels for one person is a nuisance, but one label for two people is an error.

What does ONYRI Sanitize still miss in a UK legal document?

  • Numbers labelled “UTR” or “Payroll no.”. Written in full, “Payroll number” and “Unique Taxpayer Reference:” (with the colon) are caught.
  • Police collar numbers.
  • A first name alone that the name list does not know: “Tendai” stayed visible, “David” was masked.
  • Initials and a lone surname on a signature line, such as “L. Marchetti”.
  • Health facts written in words, such as “she was pregnant”.
  • Job titles and rare events that identify someone. A clinic or street named after a person can be partly masked.

How do you review the result before it leaves your desk?

  1. Read the country line, then “Detected data”. Each value has a checkbox: ticked means masked. Untick what the recipient needs, such as a date, an amount or a case number.
  2. Use “Select text” for a missed word, “Draw an area” for a signature or a stamp, and “Also mask” for a name that returns often.
  3. Click a mask and choose “Don’t mask” if it hides too much.
  4. Read every page of the preview, then download.

The download is a new, flattened file with no text layer, so nothing can be copied from under a mask. Comments and document properties are not carried over. The file name gets “-anonymized” added, so rename it if the name is sensitive.

If a tribunal has made an anonymity order, its wording decides what must be hidden, and the tool does not know your orders. Presidential Guidance Note 2 of 2022 covers the First-tier Tribunal (Immigration and Asylum Chamber). It says such an order may bind everyone, not only the parties. A breach may be punished as contempt of court, which means disobeying the court.

What can you do without an account, and when do you need a plan?

Without an account, the tool takes one document a day from your connection, up to 5 pages, with the Black marker only. A bundle or any file over 5 pages needs the 7-day pass (€5.90, one payment, no renewal) or Pro (€9.90 a month).

Frequently asked questions

Does ONYRI Sanitize replace the fee earner’s review?

No. Detection is not exhaustive, the UK rules depend on the document being recognised as British, and some things need a hand. The check stays with you.

Is my document uploaded anywhere?

No. Reading, detection and masking run in your browser. With an account, the server receives only counters, statistics without content and a coded token of the file name. It never receives the file, its text or a detected value.

Can it read a scanned letter or a Word file?

Yes, within limits. A scan is read by OCR (text recognition) on your device, which is slower and depends on the quality. Handwriting is not read. A Word file comes back as a flattened PDF. A page it cannot read blocks the export until you mask it by hand and tick the review box.

Can I process a whole bundle at once?

The tool takes one document at a time. If you drop several files, only the first is taken, so a bundle must be one file. Anything over 5 pages needs the 7-day pass or Pro.

Sources & references

Mask a document without uploading it

ONYRI Sanitize finds names, identifiers, bank details and secrets in a PDF, a Word file or a scan, and masks them in your browser. You check the preview, then download a flattened copy.