EU AI Act — Staffing & Recruitment Evidence
Employment and workers management AI is Annex III high-risk when used for recruitment, assignment, or performance evaluation. Staffing firms and ATS vendors must evidence automatic logging at the decision boundary — not rely on vendor dashboards or mutable HR system logs.
Who is in scope
Deployers and providers of AI for CV screening, candidate ranking, interview analysis, worker assignment, and performance monitoring fall under Annex III category 4 (employment). Article 12 automatic logging and Article 26 deployer duties apply — including retention and human oversight evidence.
| Workflow step | Audit expectation | ai_act.* event |
|---|---|---|
| Candidate application ingested | Automatic record | ai_act.input_received |
| Model scores / ranks CV | Session + model version | ai_act.inference_started → output_generated |
| RAG over job description | Retrieval trace | ai_act.context_retrieved |
| Recruiter reviews shortlist | Human oversight | ai_act.human_review_completed |
| Candidate rejected / advanced | Downstream action | ai_act.action_executed |
| Serious bias incident | Post-market signal | ai_act.serious_incident_reported |
How the EU AI Act affects hiring AI
Employment AI sits squarely in Annex III. Providers must technically enable automatic lifetime logging (Article 12). Deployers retain records (Article 26), monitor operation, and evidence human oversight. Hiring chatbots can also need Article 50 transparency from August 2026.
| Situation | If high-risk rules apply | How Sigigo helps |
|---|---|---|
| CV screening / ranking ATS features | Automatic lifetime logging + retention | Instrument inference boundary; signed session chain + export |
| Interview analysis / scoring | Same Art. 12 / Art. 26 pattern | Link model output to recruiter review and hire/reject action |
| Performance / assignment AI | Workers-management Annex III path | Prove decisions and overrides without mutable HR tables alone |
| Candidate-facing chatbots | Art. 50 disclosure; Art. 12 if high-risk employment AI | Witness disclosures + optional inference events |
Article 12 requires automatic logging for high-risk employment AI — not cryptographic integrity by itself. Sigigo adds signed, exportable evidence. Field map: Article 12 guide. Primary deadline for standalone Annex III systems: December 2027.
Typical gaps in HR tech stacks
| Approach | Article 12 fit | Trust under scrutiny |
|---|---|---|
| ATS activity log | May capture events | Mutable; admins can rewrite history |
| Vendor SaaS audit UI | May satisfy retention | Verifier must trust the vendor |
| Exported CSV of scores | Incomplete session linkage | No integrity checks when challenged |
| DPIA PDF only | Process doc ≠ system log | Not automatic per-decision recording |
Privacy at the boundary
CVs and interview data are special-category adjacent in many deployments. Prefer input_hash and input_ref under Private Commitment Mode — prove the decision happened without shipping raw PII to your evidence vendor. See data sovereignty.
FAQ
- We only recommend candidates — humans always decide. Is logging still required?
- High-risk employment AI still requires automatic logging over the system lifetime. Human oversight must also be evidenced — ai_act.human_review_completed linked to the same session_id.
- Does Article 50 chatbot disclosure apply to hiring chatbots?
- Yes, if candidates interact with AI. Witness disclosure with ai_act.transparency_disclosure — see Article 50 guide.
- Where is the legal staffing explainer?
- FLI staffing guide — Sigigo owns the technical evidence map on this page.
Map your ATS or staffing workflow: run the readiness quiz or contact us.
This guide supports qualification and architecture planning — it is not legal advice. Work with your legal and compliance teams on classification, DPAs, and deployment approvals. See also our Regulatory & Compliance page.