Guide

Anonymize a SIRET, a CV or a transcript before AI.

A company number, a CV, the transcript of an interview: three kinds of content that are quickly pasted into an AI. What they reveal, and how to prepare them before sending.

Updated 14 October 2026

Why these three.

Each mixes direct identifiers (names, numbers, contact details) with details that, together, let someone be recognized: an employer and a date, a place and a role. The French Bar’s guide recalls staying vigilant about identification by cross-referencing data, even pseudonymized (p. 13).

A SIRET in a prompt.

A SIRET or SIREN number identifies a French company. For a company, that is the identity of a legal entity; for a sole proprietorship, it can point to a person, all the more next to a name or an address. When in doubt, replace it. Anoni finds SIRET and SIREN numbers, IBANs, addresses, emails and phone numbers, offline, and replaces them before the copy to the AI. If the task (rephrasing a letter, summarizing a contract) does not need the number, the AI does not either.

A CV.

The name and contact details are the obvious part; employers, dates and training are less obvious, and identify a person just as much. To have a career analyzed or a profile rephrased, the AI needs neither the name nor the contact details: Anoni replaces names, addresses, emails, phone numbers and dates, and finds organizations. A photo or a signature is not text: as with a signature, draw a redaction zone over it. Finally, reread the career: an unusual one stays recognizable.

A transcript.

Interview, meeting, lecture: a transcript mixes names, places mentioned and roles, sometimes misspelled, and a misspelled name can escape detection. Reread, select a missed word by hand, add recurring names to your personal dictionary. Anoni works on text and documents, not on an audio recording: transcribe first.

The steps, for all three.

The same path works for each:

  1. Open the document in Anoni: it finds sensitive data offline.
  2. Review the detections; add what is missing by hand or to your personal dictionary.
  3. Choose: realistic pseudonyms, tokens such as “[PERSONNE_01]”, or redaction.
  4. Click “Copy for AI”, paste into the AI of your choice, then paste the reply back at the “Export” step to restore the real values.

The limits.

Automatic detection does not replace review. Pseudonymized data remains personal data, subject to the GDPR: Anoni reduces what leaves, it does not take you out of it. And it does not decide for you what may be sent to an AI.

Frequently asked questions

Is a SIRET personal data?

Not always. For a company, it is the identity of a business. For a sole proprietorship, it can point to a person, especially next to a name or an address. When in doubt, replace it.

What about the photo and signature on a CV?

They are not text. Draw a redaction zone over them: in the exported PDF, it becomes a black bar.

Does Anoni process an audio recording?

No. It works on documents and text. For a recorded interview, transcribe it first, then prepare the transcript.

Can I then use any AI?

The prepared text can be pasted into the AI of your choice. It is for you to judge whether the tool suits the document’s sensitivity: pseudonymizing reduces what leaves, without making the data anonymous.

Reference: the French Bar (CNB) guide on ethics and AI (March 2026), in French and quoted in our translation, cited only for vigilance about identification by cross-referencing. This reference is neither a certification nor a validation of Anoni. The full method: anonymize a document before AI.