How to protect a CV before analysing it with AI
A CV combines identity, contact details, career history and indirect signals such as dates, institutions and locations. Before using it with AI, restrict content to the purpose and separate professional assessment from details that should not influence it.
1. Decide which question the AI should answer
Summarising experience is different from ranking applicants or drafting an interview. Define the use and criteria before uploading files. A vague purpose encourages unnecessary disclosure and makes output harder to assess.
- Write down job-related criteria.
- Exclude fields that do not contribute to those criteria.
- Avoid solely automated decisions without the appropriate framework.
2. Separate contact details and unnecessary signals
Name, photograph, address, email, phone, profiles, signature and identity documents are often unnecessary for summarising skills. Review nationality, birth date, marital status, disability, affiliations and categories that may invite bias.
- Replace the name with CANDIDATE_1.
- Generalise locations and dates when region or experience is sufficient.
- Keep qualifications only when relevant.
3. Preserve useful professional context
A useful copy retains roles, skills, responsibilities and outcomes. Do not replace every employer with one label if that obscures progression; use EMPLOYER_1 and EMPLOYER_2, and retain the sector where relevant.
- Check consistent placeholders.
- Ensure replacements do not alter periods or achievements.
- Separate interview notes and CVs when they serve different purposes.
4. Review the output, provider and decision
Minimising the file does not correct model bias or validate inferences. Review the response, compare it with human criteria and document the tool's role. Check the approved provider's terms.
- Do not make an automated score the final decision without the right framework.
- Check for invented facts or unsupported inferences.
- Keep a human route to correct the result.
Limitations to keep in mind
- Hiding a name does not prevent unusual studies, employers or dates from identifying a person.
- Privalume does not identify every sensitive category or assess bias.
- Preparing the CV does not replace employment, transparency and data protection duties.