EU AI Act checklist for HR and recruitment AI tools
Recruitment and candidate evaluation are among the use cases most likely to raise Annex III questions. Start by documenting intended purpose, human oversight, vendor evidence, candidate notices, and the decision workflow.
Who should use this checklist
Use this page if your team screens CVs, ranks candidates, generates interview scores, recommends hiring decisions, or buys software that influences recruitment, selection, or promotion decisions. The point is not to declare every HR tool high-risk automatically. The point is to gather the evidence needed to decide safely.
- Employers deploying third-party HR AI tools.
- Recruitment platforms selling into the EU.
- People teams using AI to summarise, rank, score, or shortlist applicants.
- Vendors answering procurement and legal review questions.
Evidence checklist
The Commission’s May 2026 draft high-risk classification guidelines make practical examples especially important. Treat those examples as classification support, not as a replacement for your own role and intended-purpose analysis.
| Area | Evidence to collect | Why it matters |
|---|---|---|
| Intended purpose | Job descriptions, workflow screenshots, decision rules, user instructions | Recruitment and candidate-evaluation uses are one of the clearest Annex III watch areas. |
| Human oversight | Reviewer roles, escalation paths, override logs, training notes | You need to show how people understand and supervise the system instead of rubber-stamping it. |
| Data and testing | Vendor validation notes, bias checks, data limits, monitoring records | Candidate-facing tools create fairness, accuracy, and documentation questions early. |
| Transparency | Candidate notices, privacy copy, chatbot or ranking explanations | Applicants should not discover AI involvement only after a decision is made. |
Fast first pass
The first useful review can be simple: describe the tool, identify who provides it and who deploys it, map where AI output enters the hiring process, collect vendor documentation, and capture screenshots of candidate-facing notices. If confidence remains low, run a deeper evidence scan before procurement or launch.
- Do not rely only on marketing pages.
- Ask the vendor for model/system documentation and limitation statements.
- Record whether humans can override, ignore, or challenge the AI output.
- Keep candidate-facing copy and workflow screenshots as evidence.
FAQ
Is every HR AI tool automatically high-risk?
No. Classification depends on intended purpose and use context. But recruitment, selection, and candidate-evaluation tools deserve early evidence collection because they are close to Annex III territory.
Should we wait for final high-risk guidance?
No. The draft guidance is useful context, but current-law planning should continue. Use it to improve classification questions and evidence requests.
Turn this reading into an actionable report
Use the free scanner to map your likely role, detect likely obligations, and see which evidence is missing.