Anonymize insurance claims for AI
To anonymize insurance claims for AI, you start from the documents a claims handler or a broker works with every day: a claim notification in PDF, a loss adjuster's report, policy schedules in Word, a loss history spreadsheet in Excel. Each one names the policyholder, the third party, the adjuster, and carries identifiers that are enough to find a file: claim number, policy number, license plate. 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 framework of the GDPR and trade secrets, and the AI uses that follow.
An insurer's files and what they contain that is sensitive
An insurance file mixes contractual documents, third-party reports and exchanges with the policyholder. Here are the ones we find at every insurer and broker, with what each one reveals.
Claim notifications and claim files (PDF, Word)
Claim number, identity of the policyholder and third parties, circumstances, license plates, accident report, injuries described.
Loss adjuster reports (PDF)
Assessment number, the adjuster's name, address of the property, cost estimate, remarks on the policyholder's lifestyle.
Policies and policy schedules (PDF, Word)
Policy number, sums insured, declared history, no-claims bonus, IBAN for direct debit, named beneficiaries.
Quotes and premium notices (PDF)
Quote and premium notice numbers, negotiated rate, family and professional situation.
Portfolio and loss history (Excel, CSV)
One row per contract or claim: member number, name, premium, cost of the claim, cause, opening date.
Income protection and health files (PDF)
Medical questionnaires, sick leave, disability, death: health data subject to stricter rules.
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 insurer profile enabled. The local analysis highlights people, addresses, dates, license plates, IBANs, claim, policy and member numbers.
- You untick false positives (the name of a cover, a car brand) and add what is missing, all occurrences at once.
- Each item becomes a stable alias (ASSURE_001 for a policyholder, SINISTRE_001 for a claim, POLICE_002; the alias labels are in French), identical from one file to the next for the same policyholder.
- In Word and PDF, redacted data is highlighted in black, an option enabled by default: you see at a glance what was removed. In Excel and CSV, columns and formulas are preserved.
- 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 batch of claims, 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 insurer profile adds contract and claim identifiers to the general detections. Here is the exact list from the software:
| Field | Where you find it |
|---|---|
| Claim number | Notifications, letters, reports, tracking sheets |
| Policy number | Policy schedules, endorsements, premium notices |
| Member number | Group contracts, income protection, health |
| Quote number | Proposals, comparisons, follow-ups |
| Premium notice number | Premium calls, renewal statements |
| Assessment number | Adjuster reports, notices of inspection, estimates |
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 company or brokerage can be added by keyword (see anonymization by profession).
Trade secrets, GDPR and the insurer's obligations
The GDPR applies to all policyholder data, with stricter rules for the health data of income protection contracts. French Law No. 2018-670 of 30 July 2018 on the protection of trade secrets protects pricing terms, the loss history of a portfolio or clauses negotiated with a key account. Finally, Article L113-2 of the French Insurance Code governs how a claim is notified, which makes the file a contractual document to be kept faithfully.
Two consequences for the handler's files. With offline software, the named file stays inside the company or brokerage: you keep control over your subcontractors and your trade secrets. Health information in income protection files is masked, beyond simple pseudonymization. Biffer prepares, you validate: the handler reviews the preview and has the final word on what is shared.
AI use cases once the files are redacted
A redacted file keeps its logic: ASSURE_001 reports claim SINISTRE_001 under policy POLICE_002. The AI follows the thread without knowing anyone.
- Summarize a thirty-page adjuster report before the coverage decision.
- Draft a reply to a complaint, from the file and the general terms.
- Extract the covers, deductibles and exclusions of a policy for a comparison.
- Spot inconsistencies between a notification, an accident report and an assessment.
The redacted file can be pasted into ChatGPT, Claude or Gemini; the page anonymize a document details the policies per data type. 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
Are license plates detected in a motor claim file?
Yes. Biffer recognizes French plates in the current SIV format (AA-123-AA) and the older FNI format, in the body of an accident report, an adjuster's report or a claims spreadsheet, and replaces them with a stable alias (PLAQUE_001). A plate is an indirect identifier: it leads to the registration certificate, and so to the holder. It is redacted just like the name. If a technical analysis needs to tell vehicles apart, the alias is enough: VEHICULE_001 hits VEHICULE_002 stays perfectly readable, and the mapping table restores the real plates on the way back.
How do you handle health data in an income protection file?
With the "mask" policy rather than "pseudonymize". A name becomes ASSURE_001 and stays useful for following the person through the document. A condition, a period of sick leave, a disability rate are health data within the meaning of Article 9 of the GDPR: they disappear from the file you send, beyond a simple change of label. For this, Biffer offers masking per type and the custom keyword field, which spots your internal labels (reason, diagnosis, long-term condition). You combine policies in a single file, then check the preview before export.
Can you redact a batch of fifty claim files at once?
Yes, with the queue. You drop the files (PDF, Word, Excel, CSV) and choose between reviewed mode, where you check the preview of each file, and automatic mode, which applies your policies to the whole series. For similar files, automatic mode saves time; for a first batch, we recommend reviewed mode. Aliases are shared across the files in the batch: the same policyholder or adjuster gets the same alias everywhere, which allows a consistent portfolio analysis by the AI.
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