Document anonymization for professionals: your files, ready to share publicly or with an AI tool.
Redact your docs, pdf, jpg, png, xlsx, csv, txt, docx, tiff, webp, md, xlsm
Biffer is document anonymization software: it removes personal and sensitive data from your documents before you share them publicly or use them with ChatGPT, Claude, Copilot, Gemini, Mistral or any other AI/LLM.
100% offline · Free: 5 files per month · Your files stay on your computer
Biffer downloads
Document
Attendees: Marie Martin and Paul Lefebvre
Contact: marie.martin@exemple.fr
Phone: 06 12 34 56 78
File no. 2024-0157
SIREN 552 100 554
Address: 12 rue de la Paix, 75002 Paris
Decision: follow up with the client next week.
Download the docSample PDF, before redaction
Redacted document
Download the docThe same PDF, redacted by Biffer
Built for professions that handle sensitive data
Redact, black out, anonymize, de-identify, depersonalize, pseudonymize, mask, obscure, expunge, sanitize: whatever word your profession uses, the need is the same, removing from a document whatever identifies a person or a case before you share it. A lawyer, an accountant, a doctor, a payroll officer, a local government officer or a teacher each have their own files and their own identifiers. Biffer knows eighteen professions and, for each one, which fields to look for. Click a profession to see what the software finds in your documents.
In every case, the software detects last names, first names, emails, phone numbers, addresses, French company numbers (SIREN and SIRET), French social security numbers, IBANs, license plates and dates. If an identifier specific to your firm or department is missing, you declare it in a moment with a keyword, and it is redacted like the others.
What the AI Act says about confidential data and AI
The EU regulation on artificial intelligence (Regulation (EU) 2024/1689, known as the AI Act) entered into force on 1 August 2024 and applies in stages until 2027. It does not replace the GDPR: it classifies AI uses by level of risk and sets obligations for providers and for "deployers", meaning you, as soon as you use AI in a professional setting. The confidential documents you hand to a tool remain governed by the GDPR and by professional secrecy. Choose your situation below and the rules that apply are shown alongside. We are not lawyers: this overview points to the official texts and does not replace legal advice.
What you do
What the AI Act says
I summarize or rewrite a client document with ChatGPT, Claude or Copilot
Minimal risk- AI Actyou are a "deployer" of a general-purpose AI system. For this everyday use, the regulation imposes no specific formality, but since 2 February 2025 Article 4 requires you to ensure the AI literacy of your staff (understanding what the tool does with data, and its limits).
- GDPRthe AI provider becomes a recipient of the personal data in the document. You need a legal basis, compliance with data minimization (Article 5) and a contract governing the processor (Article 28); any transfer outside the European Union must be covered.
- Professional secrecyfor a lawyer, a doctor, a notary or an accountant in France, passing the identifying content of a file to a third party engages professional secrecy (Article 226-13 of the French Criminal Code).
- In practiceredact the document before uploading it; the AI works on aliases and the answer is restored on your computer. The issue disappears along with the data.
I screen résumés or assess candidates and employees with AI
High risk- AI Actemployment and worker management are listed in Annex III: recruitment, selection, evaluation, promotion and task allocation are high-risk uses. As a deployer, from 2 August 2026 Article 26 requires human oversight, use in line with the provider's instructions, relevant input data, retention of logs and informing employees and their representatives before the system is put into service.
- Prohibited(Article 5, since 2 February 2025): emotion recognition in the workplace, except for medical or safety reasons.
- GDPRno fully automated decision producing legal effects without safeguards (Article 22); candidates must be informed.
- In practicea redacted applicant spreadsheet (names, emails, phone numbers and application numbers replaced by aliases) can be analyzed without exposing anyone; the decision stays human.
I use AI on patient or beneficiary files
Sensitive data- AI Acta system that contributes to decisions on access to healthcare or essential social benefits falls under Annex III (high risk); a tool built into a medical device falls under Annex I. In both cases the provider has stricter obligations, and you must ensure human oversight and inform the people concerned (Article 26).
- GDPRhealth data is sensitive data (Article 9), processed only by way of exception; in France, hosting health data is also regulated.
- Medical secrecyArticle L1110-4 of the French Public Health Code.
- In practiceBiffer replaces patient identifiers, RPPS numbers (French health professional IDs), admission numbers, names and dates; a redacted report can be summarized without the tool being able to identify the patient.
I publish or distribute text, images or voice produced by AI
Transparency- AI ActArticle 50 (applicable from 2 August 2026) requires content generated or manipulated by AI to be disclosed: deepfakes (images, audio, video) must be labeled, and AI-generated text published to inform the public on matters of public interest must state its origin, unless it has undergone human review and editorial responsibility.
- Providerssystems must mark their outputs in a detectable way (machine-readable marking).
- Public releasebefore anything is published, the personal data in the source document must have been removed, this time under the GDPR.
- In practiceredacting the source document prevents a name, an address or a case number from ending up in the generated text, and then in print.
I train or fine-tune a model on my own files
Provider- AI Actyou become a provider. For a high-risk system, Article 10 requires governance of training data (relevance, representativeness, bias examination, provenance); for a general-purpose model, Article 53 requires technical documentation, a copyright compliance policy and a public summary of the training data (obligations applicable since 2 August 2025).
- GDPRthe French data protection authority (CNIL) recommends building training sets that are minimized, anonymized or pseudonymized, with a defined purpose; the EDPB (Opinion 28/2024) points out that whether a model is anonymous must be assessed case by case.
- In practicerun each file through Biffer before adding it to a training set; keep the mapping table separately, or do without it for permanent anonymization.
I have creditworthiness, an insurance risk or a public benefit application assessed
High risk- AI Actassessing the creditworthiness of individuals, pricing life and health insurance, and granting essential public benefits are high-risk uses (Annex III). Deployer obligations from 2 August 2026: human oversight, informing the people concerned, logs, and a fundamental rights impact assessment (Article 27) for public bodies and for banking and insurance institutions.
- GDPRthe right not to be subject to a fully automated decision (Article 22), and information about the logic involved.
- In practiceredacted supporting documents (statements, pay slips, tax notices) keep what the analysis needs (amounts, dates, categories) without the identity or account numbers.
Uses the AI Act prohibits, whatever the sector
Prohibited- Since 2 February 2025 (Article 5), prohibited practices include: subliminal manipulation or exploitation of vulnerabilities, general social scoring, assessing the risk that a person will commit an offense based solely on profiling, building facial recognition databases through untargeted scraping of images, emotion recognition in the workplace and in education, biometric categorization inferring sensitive data, and real-time remote biometric identification by law enforcement (with narrow exceptions).
- Penaltiesup to €35 million or 7% of worldwide annual turnover for prohibited practices.
- In practiceno anonymization makes a prohibited use lawful; Biffer serves permitted uses by removing what has no reason to circulate.
Sources: Regulation (EU) 2024/1689 on EUR-Lex, European Commission page on the AI Act. Timeline: prohibitions and AI literacy since 2 February 2025, general-purpose models since 2 August 2025, Annex III high-risk systems from 2 August 2026, Annex I regulated products in 2027.
Who uses Biffer
Biffer is used every day by professionals bound by secrecy or confidentiality: they want the benefits of AI assistants without handing over the names, numbers and addresses of their clients, patients or employees. The testimonials below illustrate these uses by profession; the first names and organizations are fictitious until our first users agree to be quoted. Each situation links to the page for that profession, with the list of fields the software detects.
For months I have been summarizing opposing briefs and drafting letter templates with an AI, always keeping my clients' names to myself. Biffer replaces the parties, the RG case numbers and the addresses in a few seconds, and I restore everything in one click in the answer.
law firm, Lyon · see the files of a lawyer
Our clients' balance sheets and FEC files (the French standard accounting export) go through Biffer before any assisted analysis. VAT numbers, account numbers and invoices become aliases, the rest of the Excel file stays intact. All my staff have a license on their computer.
accounting firm, Nantes · see the files of an accountant
To dictate and then rephrase a consultation report, I wanted the patient's file to stay with me. Biffer removes names, dates of birth, social security numbers and RPPS numbers first, and everything stays on my computer.
medical practice, Bordeaux · see the files of a general practitioner
Sorting applications or preparing annual reviews with an AI, yes, but without names, employee IDs or social security numbers. The black highlighting shows at a glance what has been removed from the document.
industrial SME, Lille · see the files of an HR manager
Deeds and land registry references are pseudonymized in one pass, including in scanned PDFs thanks to OCR. The encrypted mapping table lets us recover the real values when the reviewed deed comes back.
notary's office, Rennes · see the files of a notary's clerk
Training agreements, trainee lists, attendance sheets: I redact everything before asking an AI to write a summary or a certificate. Five free files a month were enough at first, then I bought the unlimited license.
training organization, Lyon · see the files of an independent trainer
Illustrative testimonials. Do you use Biffer and would like to share your experience? Write to us at contact@biffer.fr.
How Biffer, the offline anonymizer, works
Biffer installs on your Mac or PC and works directly in your computer's memory: it reads the document, finds the personal data, replaces it with aliases and encrypts the mapping table, all on your machine. Your files, your clients and your patients stay with you, and you alone hold the key that brings back the real names. You can then use ChatGPT, Claude, Copilot or Gemini with peace of mind: the AI receives a usable document, and you get back a deliverable that Biffer restores with the original names in one click. You work wherever you like, on the train, at a client's office or on the road, even without a network. A single connection is needed, the first time you open the software, to receive your key by email. Try it below: click Redact, then Analyze, to see what ChatGPT receives, get the deliverable back and restore it.
Your computer
Consultation report of March 12, 2024
Patient: Jean Martin, born on 04/07/1968
Address: 12 rue de la Paix, 75002 Paris
Phone: 06 12 34 56 78, email: j.martin@exemple.fr
Social security no.: 1 68 07 75 123 456 78
Reason: lower back pain for three weeks.
History: surgery in 2019, Sainte-Anne clinic.
Conclusion: rest, physiotherapy, follow-up on April 2.
Referring doctor: Dr Sophie Leroy, RPPS 10001234567
File no. 2024-0157, employer: Transports Durand SAS
The document scrolls on your computer, and stays there.
ChatGPT
Waiting for a document.
Receives only what you choose to send it.
Restored deliverable
ChatGPT deliverable
This approach gives concrete guarantees to a professional bound by secrecy: anonymization happens under your sole control, so this step does not require a data processing agreement within the meaning of Article 28 of the GDPR; your files exist only on your side, and what you process is known to you alone. Our offline software page explains what local processing brings you. Biffer is free for five files per month; the unlimited plan costs €10 excl. VAT (€12 incl. VAT) per computer, paid once, for life for the version purchased: see our pricing.
You have the final say. Biffer prepares, you approve: the highlighted preview shows you every detection, and you adjust whatever needs it in one click before export. On an image or a scanned PDF, redaction blacks out the pixels permanently, the guarantee that the information is really gone from the file. Biffer saves you most of the time and leaves the decision to you.
Carbon cost of anonymization: offline versus online
Anonymizing a document on your own computer takes very little energy. Biffer finds personal data with rules, not with an AI model, then masks, blacks out or pseudonymizes it: under a second for a page of text, a few seconds for a scan. Nothing goes to a data center: no network transfer, no remote storage, no computation on water-cooled graphics processors. An online service that redacts or de-identifies your files with generative AI does the opposite: it receives every page on its servers and has a large language model read it. Figures published by Google, OpenAI, Mistral AI and independent researchers put a single request between 0.24 and 1.2 Wh, that is 5 to 27 times the energy we count for redacting one page locally. Offline redaction is therefore also a choice of digital sobriety: less electricity, less water, less CO2e, and personal data that stays on your computer, as the GDPR encourages you to do. Our offline software page details the other benefits of local anonymization. Run the numbers for your own volume: we chose cautious assumptions that work against Biffer, detailed below the simulator.
That is 1,200 pages per year. Annual results.
Biffer, offline and without AI
Electricity
0.053 kWh
laptop at 40 W, 4 s per page at most
3.6 smartphone charges
Water
0.23 L
power plant water counted generously
1.2 glasses of water
Carbon
0.0027 kg CO2e
French electricity, 50 g CO2e/kWh
19 m by petrol car
Biffer takes the trophyLess electricity, less water and less carbon, on all three counts.
Online service with AI
Electricity
0.3 kWh
up to 1.6 kWh (high estimate)
20 smartphone charges
Water
1.63 L
up to 54 L (high estimate)
8.2 glasses of water
Carbon
0.043 kg CO2e
up to 1.43 kg CO2e (high estimate)
302 m by petrol car
Online AI, a carbon tractorIts servers run at full throttle and emit far more CO2e for the same task.
The gap in favor of offline anonymization
Over a year, you avoid at least 0.25 kWh, 1.4 L of water and 0.04 kg CO2e, and up to 1.54 kWh, 53.8 L and 1.43 kg CO2e under the high estimate.
5.7 to 30 times less electricity16 to 537 times less CO2e
Method. Each column adds up, page by page, electricity, water and emissions over twelve months. The right-hand column shows the low estimate, the most favorable to the online service; the high estimate appears in small print. Our assumptions per page:
- Biffer, offline: a 40 W laptop running for 4 seconds, the case of a scan read by OCR (a page of text takes under a second), or 0.044 Wh. French electricity at 50 g CO2e/kWh (46 g in ADEME's Base Empreinte, 2025 average mix, rounded up). Water: 4.35 L/kWh, the off-site factor Jegham et al. apply to data centers, counted generously for your computer.
- Online, low estimate: a single request per page, at the median Google published for Gemini (0.24 Wh, 0.26 mL of water, 0.03 g CO2e, measured in May 2025). OpenAI gives a similar order of magnitude for ChatGPT: 0.34 Wh and about 0.32 mL per request (June 2025). As for Biffer, we add the water used by power plants, at the same factor.
- Online, high estimate: 1.2 Wh for a request of 1,000 input tokens and 1,000 output tokens with GPT-4o (Jegham et al., 2025), an A4 page of text counting about 500 to 800 tokens; 45 mL of water and 1.14 g CO2e for a 400-token response, according to Mistral AI's life cycle analysis (July 2025).
- Network transfer: 0.5 MB per uploaded page, at 0.025 kWh/GB (low) or 0.23 kWh/GB (high), the range cited by the IEA; global electricity at 458 g CO2e/kWh in 2025 (Ember). Remote storage is not counted.
- What we leave out: model training, the manufacturing of servers and of your computer, and multiple requests. A long document is often split into several requests, and every retry adds up: the real gap is therefore probably wider.
Equivalents: a smartphone charge of about 15 Wh, a 20 cl glass of water, a petrol car at 142 g CO2e/km (Impact CO2, ADEME). Sources: Google Cloud, environmental impact of Gemini inference (August 2025), Sam Altman, The Gentle Singularity (June 10, 2025), Mistral AI, life cycle analysis (July 22, 2025), Jegham et al., How Hungry is AI? (arXiv, 2025), IEA, the carbon footprint of streaming video (2020), Ember, Global Electricity Review 2026, French electricity mix, ADEME Base Empreinte via Nos Gestes Climat.
Frequently asked questions
Does Biffer really work without an internet connection?
Yes. The application installs on your computer and the whole analysis happens locally: data detection, rewriting the document, encrypting the mapping table. You can disconnect from the network before starting a job and the result will be identical. The only time a connection is needed is the first time you open the software: you enter your email address in the software, click the confirmation link you receive, and Biffer receives the key tied to your address and your computer. This happens once per machine; after that the key is checked offline and the free plan counter is kept on your disk.
Which document formats can I redact?
Biffer handles PDFs (text rewritten in place, layout preserved), Word documents (.docx), Excel workbooks (.xlsx), CSV files and text files. In Word and Excel, formatting, headers, comments and rich text are preserved; metadata that names people is cleared. Images and scanned PDFs, made of a picture of each page, are read by built-in character recognition, offline; the data found is redacted by blacking out the pixels, a permanent redaction that guarantees the information is gone from the file.
Can I get the real values back after using AI?
Yes, if you kept the mapping table when you redacted the document. It links each alias (PERSONNE_001, EMAIL_002) to the original value and stays on your computer, encrypted with a password or in plain text, as you choose. Paste the AI answer into a document, open it in Biffer, click Unredact and point to the table: Biffer puts the names, emails and numbers back in place, even if the text has been rephrased. The table is the only key to restoration: by deleting it, you get permanent anonymization, a valuable guarantee when that is what you want.
What happens after the five free files of the month?
The software stays installed, with all your settings, and you can process new files again on the first day of the following month, when the counter, kept on your computer, resets to zero. You keep every feature, with no time limit. To go beyond five files per month, Biffer for Mac or Biffer for PC costs €10 excl. VAT (€12 incl. VAT), paid once, a lifetime license for the version purchased, one license per computer, not transferable to another machine. The new key, delivered through the same email sign-in in the software, removes the limit immediately; it is then checked locally, offline. The details are on the pricing page.
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