Anonymize medical records for AI
To anonymize medical records for AI, you work on the documents that come out of the practice software or the secure health messaging system: a clinical note in Word, a hospital discharge report in PDF, a prescription, a patient list exported as CSV. Each one names the patient, their date of birth, their address, their hospital patient ID (IPP), their admission number, and the doctor signs with an RPPS number (the French national register of health professionals). Biffer recognizes them on your computer, with no connection, replaces them with stable aliases and highlights them in black in the output file. The medical content itself stays intact.
On this page we go through these files one by one, how to redact them, the framework of medical secrecy and the AI uses that follow.
A medical practice's files and what they contain that is sensitive
A patient record is made of documents produced by the doctor, the hospital, the laboratory and the national health insurance. Here are the ones we find everywhere, with what each one reveals.
Clinical notes and patient records (Word, text)
Identity, date of birth, address, family history, occupation, social situation, name of the regular doctor.
Discharge and operative reports (PDF)
Patient ID (IPP), admission number, admission and discharge dates, department, names of the surgeons, the facility's FINESS number (French register of health facilities).
Prescriptions and sick notes (PDF)
Prescription or sick note number, the prescriber's RPPS and ADELI numbers, the patient's French social security number, the employer named on the sick note.
Test results and letters to colleagues (PDF, Word)
Names of the patient and of both practitioners, medical council registration number, diagnostic conclusions, direct emails and phone numbers.
Exports from practice software (Excel, CSV)
Patient lists with dates of birth, social security numbers, health insurance member numbers, visit dates and reasons.
Care forms and statements (PDF)
Supplementary health insurance (mutuelle) member number, coded procedures, amounts, IBAN for reimbursements.
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 doctor profile enabled. The local analysis highlights people, addresses, dates, French social security numbers (check key verified), RPPS, ADELI, patient IDs, admission and prescription numbers.
- You untick false positives (the name of a molecule, a clinical scale) and add what is missing, all occurrences at once.
- Each item becomes a stable alias (PATIENT_001, RPPS_001, SEJOUR_004 for an admission; the alias labels are in French), identical from one document to the next for the same patient.
- 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 series of reports, 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 doctor profile adds health identifiers to the general detections. Here is the exact list from the software:
| Field | Where you find it |
|---|---|
| RPPS number, ADELI number | Prescriptions, reports, signed letters |
| FINESS number | Facility headers, care forms |
| Patient ID (IPP) | Hospital records, labels, results |
| Admission number | Discharge reports, invoices |
| Medical council registration number | Letters, nameplates, council documents |
| Sick note number | Sick notes, extensions |
| Prescription number | Prescriptions, renewals |
| Mutuelle member number | Care forms, statements, coverage approvals |
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 practice can be added by keyword (see anonymization by profession).
Medical secrecy: Article L1110-4
Article L1110-4 of the French Public Health Code guarantees every person in care the secrecy of the information that concerns them. It restricts sharing to the professionals involved in the care, for strictly necessary information. Patients' rights over their records are summarized on service-public.fr (in French). Health data is also a special category under Article 9 of the GDPR.
Two consequences for the practice's files. A redacted report can be submitted to an online AI service, outside the circle of care, while staying within the bounds of secrecy. And the name is only one identifier among others: a patient ID, a date of birth, an admission number and a department are enough to find a patient, and the doctor profile handles them all. Biffer prepares, you validate: the professional reviews the preview and has the final word on what is shared.
AI use cases once the files are redacted
A redacted report keeps all its clinical value: PATIENT_001, aged 58 if you keep the age, their history, treatment and results. Only the identity is gone.
- Summarize several years of records before a complex consultation.
- Rephrase a letter to a colleague or an explanation for the patient.
- Extract history and treatments from a series of notes into a follow-up table.
- Prepare a clinical case for teaching or a team meeting, with no possible re-identification.
The redacted file can be pasted into ChatGPT, Claude or Gemini; the page personal data and AI recalls the special rules for health data. 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
Should the date of birth and the dates of care be redacted?
The date of birth, yes: combined with a department and a procedure date, it identifies a patient with certainty. Biffer detects all dates and lets you choose the policy per type: mask, which replaces the value with [DATE], pseudonymize with an alias, or keep. In practice, many practitioners mask the date of birth and give the age in the text instead, and keep the dates of care that the AI needs to rebuild a timeline. The choice is made in the preview, detection by detection or for all occurrences of a type, before export.
How do you handle a CSV or Excel export from practice software?
You drop the file as it is. Biffer reads each column, spots names, dates of birth, French social security numbers (whose check key it verifies), patient IDs, mutuelle member numbers and addresses, then shows the preview with each detection highlighted by type. The columns for reasons, procedures and amounts stay readable, and the file structure (separator, encoding, Excel formulas) is preserved. The output file can be analyzed by a spreadsheet or by an AI for patient statistics, with every row protected from re-identification.
Should the doctor's own RPPS number be redacted?
It depends on the use. The RPPS and ADELI numbers are public identifiers of the professional, but they link the document to a practice, and so to a patient base and a place, which reduces the patient's anonymity. For a file sent to an online AI, we advise redacting them: the doctor profile recognizes them and replaces them with RPPS_001, with no effect on the clinical content. For a clinical case shared between identified colleagues, the "keep" policy applies to that data type alone, while the patient's identifiers stay redacted.
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