Real human review and DPIA in AI-assisted hiring
Already applicable
- When it relates
- If you use AI to read, score or rank CVs or evaluate candidates or employees.
- What the rule provides
- Apply data protection by design and assess risks to candidates and workers.
- Carry out a DPIA before using the tool if processing may be high-risk.
- Give clear information about the tool's role in the assessment.
- Ensure the decision-maker can critically assess the score and make their own decision; simply accepting the output is not oversight.
- Good practice: extend human review to candidates the tool rejects, not only those who pass the filter.
- Typical evidence
- DPIA of the hiring system.
- Human review procedure including samples of rejected candidates.
- Candidate notice.
- Related base documents
- Supporting documents
- Verification
- Verified in the official source on November 2, 2026Checked against the AEPD's official press release. Correction: the research dated the warning Oct 7, 2026; the AEPD release is dated Sep 23, 2026. The release does not expressly mention rejected candidates: that point comes from commentators and is presented as good practice.