Onyrisanitize
All articles
Guide6 min read

Anonymize HR data before you hand it to AI

Résumés, payslips, reviews: HR data is among the most sensitive you hold. How to anonymize it before ChatGPT, Claude or Gemini without losing usefulness.

By Alexis de ONYRI

HR data — résumés, payslips, performance reviews, sick leave — is among the most sensitive an organization holds, and some of it falls under the GDPR's special categories. Before asking AI to summarize a review or compare applications, replace names, salaries, addresses, social-security numbers and health data with tokens. The model handles the task; no identifying data leaves the browser.

Why HR data deserves special handling

An employee file concentrates information that, alone or combined, identifies a person and reveals their private life. Some of it is a “special category” under the GDPR: health, union membership, origin. Processing it is strictly framed — and pasting such a file into a consumer assistant exposes it to a third party with no framework.

  • Health data: sick leave, workplace accommodations, medical restrictions.
  • Compensation: salaries, bonuses, family and tax situation.
  • Direct identity: name, address, social-security number, contact details.
  • Assessments: performance reviews, interview notes, disciplinary data.
Diagram: an employee record whose identity, salary and health data are masked by tokens, passing through an anonymization gate before reaching the AI.
The HR record passes through an anonymization gate: only masked text reaches the AI, the mapping stays local.

The HR fields to mask first

  • Identity and contact details: first name, last name, address, email, phone.
  • Social-security number and administrative identifiers.
  • Compensation: salary amounts, bonuses, payroll elements.
  • Health: leave, restrictions, any medical element.
  • Named assessments and disciplinary data.

The right reflex: anonymize, don't give up

  1. 1Detection: the engine spots identity, salary, social-security number and health data.
  2. 2Tokenization: each element becomes a neutral token, kept in local memory.
  3. 3Sending: only the anonymized text goes to the AI — the HR task stays doable.
  4. 4Restoration: the answer is de-tokenized in your browser, ready to use.

ONYRI Sanitize detects sensitive HR data — from the name to the salary amount and the social-security number — and restores the answer in your browser. Your HR teams gain AI's speed to summarize or structure, without ever exposing an employee.

Frequently asked questions

Can I use AI to screen résumés?
With caution. Automated screening of applications can fall under a framed use (watch for bias and for decisions affecting people). On the data side, anonymize identity and contact details before sending: AI evaluates skills, not the named person. The final decision stays human.
Are salaries sensitive data?
They aren't a “special category” in the strict GDPR sense, but they are personal data whose disclosure is highly intrusive. As such, they deserve the same anonymization reflex before any send to a third-party tool.
How do I summarize a performance review without exposing the employee?
Replace the name and identifying elements with consistent tokens, then ask for the summary. AI works on the review's content; after restoration in your browser, you get back a complete text, tied to the right person — without their identity ever leaving your machine.

Sources & references

Keep your sensitive data in your browser

ONYRI Sanitize detects and masks your sensitive data before it reaches the AI, then restores the answer — from names to API keys.

Anonymize my prompt

Read next