Anonymize notary documents for AI
To anonymize notary documents for AI, you start from the files that circulate in a French notary's office (the notaire is a civil-law notary who drafts and authenticates property sales, estates and marriage contracts): a draft deed in Word, a certified copy in PDF, a mortgage search, a loan schedule in Excel. Each one names the parties, their spouses, their heirs, and carries identifiers that point to a property as surely as a name: land registry reference, lot number, deed number. Biffer recognizes them on your computer, with no connection, replaces them with stable aliases and highlights them in black in the output file.
On this page we go through these files one by one, how to redact them, the ethical framework and the AI uses that follow.
A notary office's files and what they contain that is sensitive
A notarial file gathers documents from every source, from the signed original deed to a letter from a colleague. Here are the ones we find in every office, with what each one reveals.
Sale deeds and preliminary contracts (Word, PDF)
Full civil status of the parties, matrimonial regime, addresses, price, land registry references and lot numbers, chain of title over several owners.
Estates and gifts (Word, PDF)
Estate number, list of heirs with dates of birth, detailed assets and liabilities, bank accounts.
Mortgage searches and land registry documents (PDF)
Filing numbers, registered liens and mortgages, plots, identity of the creditors.
Loan files (PDF, Excel)
Loan number, amount, repayment schedule, the borrower's declared income, IBAN and guarantees.
Co-ownership: bylaws and general meeting minutes (PDF)
Co-ownership number, names of the co-owners and their shares, named votes, unpaid charges mentioned in the meeting.
Formalities and correspondence (Word, text)
CRPCEN number of the office and of other offices (the identifier of French notary offices), register number, exchanges with the land registration service and the banks.
Text PDFs, Word, Excel, CSV and plain text go through the same process (see redact a PDF); images and scanned PDFs go through built-in optical character recognition, offline, then are redacted by blacking out the pixels, with no possible restoration.
Redacting these files in practice: formats, aliases, black highlighting
Everything happens on your computer, with no connection (see offline software).
- You drop the file with the notary profile enabled. The local analysis highlights people, addresses, dates, IBANs, CRPCEN, deed and loan numbers, and land registry references.
- You untick false positives (the name of a town, an article of the Civil Code) and add what is missing, all occurrences at once.
- Each item becomes a stable alias (PERSONNE_001, CADASTRE_002, ACTE_001; the alias labels are in French), identical from one document to the next for the same party or the same property.
- In Word and PDF, redacted data is highlighted in black, an option enabled by default: you see at a glance what was removed. Formatting is preserved, metadata is cleared.
- You choose whether to save the mapping table, encrypted (Argon2id and AES-256-GCM) or in plain text. With it, the real names are restored in the AI's answer; without it, the redaction is irreversible.
The queue takes a complete file, appendices included, in reviewed or automatic mode. Biffer is free for five files per month, all redaction features included, after a one-time email confirmation in the software, with no password. Beyond that, Biffer for Mac or Biffer for PC costs €10 excl. VAT (€12 incl. VAT) per computer, paid once by card via Stripe: a lifetime license for the version purchased, tied to the computer, unlimited files, each major version sold separately. Every feature is included from the free plan (see pricing).
The notary profile adds the identifiers of the deed and the property to the general detections. Here is the exact list from the software:
| Field | Where you find it |
|---|---|
| CRPCEN number | Headers, letters between offices, formalities |
| Deed number | Original deeds, certified copies, register |
| Land registry reference (cadastre) | Sale deeds, mortgage searches, chain of title |
| Register number | Office register, marginal notes |
| Filing number | Land registration, mortgage searches |
| Estate number | Estate returns, affidavits of heirship, letters |
| Loan number | Loan offers, repayment schedules, loan deeds |
| Co-ownership number | Bylaws, meeting minutes, pre-sale statements |
On top of these come the fields shared by all professions (file number, reference, contract number, employee number) and the general detections: people, emails, phone numbers, addresses, French company numbers (SIREN and SIRET), French social security number, IBAN, license plates, dates. An identifier specific to your office can be added by keyword (see anonymization by profession).
The French notary's professional secrecy
Decree No. 2023-1297 of 28 December 2023 on the notaries' code of ethics states that the notary and every person under their authority are bound by a general and absolute professional secrecy, covering deeds, documents, exhibits and correspondence, electronic included. The professional rules approved by the order of 29 January 2024 complete this framework, and breach of secrecy is punished under Article 226-13 of the French Criminal Code.
Two consequences for the office's files. Property identifiers are redacted like people: a land registry reference, a lot number and a town are enough to find an owner. With offline software, processing stays in the office, from import to export. Biffer prepares, you validate: the notary reviews the preview and has the final word, as with any deed.
AI use cases once the files are redacted
A redacted deed keeps its structure and clauses: PERSONNE_001 sells to PERSONNE_002 lot LOT_003 in co-ownership COPRO_001. The AI can analyze it while the people and the property stay protected.
- Summarize co-ownership bylaws or a series of general meeting minutes before a sale.
- Check that a preliminary contract matches the draft final deed.
- Extract the easements and special conditions from an old title deed.
- Explain a clause in plain language for the letter to the client.
The redacted file can be pasted into ChatGPT, Claude or Gemini; the page redact a document explains the difference between visible redaction and pseudonymization. The answer, loaded back into Biffer with the mapping table, gets the real names back, still locally. Biffer is designed in Lyon by Romain Rissoan, its founder, trainer and consultant, at OPCADIA.
Frequently asked questions
Is a land registry reference personal data to redact?
Yes, as long as it designates a property whose owner can be identified by simply consulting the land registry or the property file. Section, plot number and town form an identifier as precise as a name. Biffer's notary profile therefore treats the land registry reference, the lot number and the co-ownership number as fields in their own right, replaced by stable aliases (CADASTRE_002, LOT_003) that stay identical in the preliminary contract, the deed and the mortgage search. If an analysis requires keeping the property description, the "keep" policy applies to that data type alone.
How do you redact a sixty-page deed with its appendices?
You drop the deed and its appendices into the queue (Word, PDF, Excel). Aliases are shared across all the files in the batch: the same party gets the same alias in the deed, the co-ownership bylaws and the repayment schedule. In reviewed mode, you check the preview of each file; in automatic mode, Biffer applies your policies to the whole series, which suits appendices of the same kind. For a first file, we recommend reviewed mode, long enough to confirm that the office's internal identifiers are properly recognized.
Is black highlighting suitable for a copy of a deed sent to a third party?
That is exactly its purpose. By default, Biffer highlights each redacted item in black in Word and PDF files: the reader sees that a passage was removed, and the content stays protected, since the original value has left the file and an alias has replaced it. This visible mark suits a working copy given to an expert or to an AI. If you prefer an ordinary-looking document, the option can be turned off before export. Processing always applies to a copy: the original deed and the certified copies stay intact in the office.
Your business documents anonymized, ready to be published or handed to an AI tool.
Biffer runs on your computer without sending a single file. Free for five files a month, unlimited for €10 excl. VAT per computer.
100% offline · Free: 5 files a month · No file sent