EU AI Act Compliance for Recruitment & HR: AI in Workforce Management

Recruitment (Art. 6 + Annex III, 4(a))

• CV screening and ranking: AI that filters, scores, or ranks job applications — HireVue, Pymetrics, any ATS with AI screening
• Automated video interviews: AI analysing facial expressions, tone of voice, or word choice to assess candidates
• Job advertisement targeting: AI deciding which candidates see which job postings (if it materially affects access to employment)
• Chatbot screening: Conversational AI that pre-screens candidates and decides who advances

Workforce Management (Art. 6 + Annex III, 4(b))

• Performance evaluation: AI systems used for promotion decisions, performance ratings, or identifying underperformers
• Task allocation: AI assigning work, shifts, or projects based on worker profiling — especially in gig economy and warehouse/logistics
• Monitoring and surveillance: AI-driven employee monitoring (keystroke tracking, email analysis, productivity scoring) when used for employment decisions
• Termination decisions: AI systems influencing who gets terminated, made redundant, or not renewed

For HR Technology Providers (Providers)

• Conformity assessment: Internal conformity assessment (no third-party audit needed for HR AI). Self-certification with technical documentation.
• Bias testing: Test for discrimination across protected characteristics — gender, age, ethnicity, disability. Document results and mitigation measures.
• Technical documentation: Training data description, model architecture, performance metrics, limitations, intended use
• Human oversight: Design the system so a human can effectively oversee and override AI decisions
• Transparency: Inform users (employers) that the system is high-risk AI. Provide clear instructions for deployment.
• EU database registration: Register the system in the EU AI Act database before placing it on the market

For Employers (Deployers)

• Fundamental rights impact assessment: Before deploying high-risk HR AI, assess impact on workers' fundamental rights (Art. 27)
• Inform workers: Employees and candidates must be told that AI is being used in decisions affecting them
• Inform works council: In EU countries with mandatory worker representation, works councils must be informed before deployment
• Human oversight: Ensure trained humans review AI recommendations before final decisions. Never fully automate hiring or termination.
• Data protection: GDPR Art. 22 already restricts automated decision-making. EU AI Act adds additional obligations.

1. Inventory all HR AI: Map every tool in your recruitment and HR stack that uses AI — ATS, video interview platforms, performance management, workforce scheduling
2. Classify risk: Which are high-risk under Annex III? Some HR tools use only basic automation (not AI) — these may not be in scope.
3. Vendor due diligence: Ask HR tech vendors: Are you compliant with the EU AI Act? Can you provide conformity documentation? What bias testing have you performed?
4. Human-in-the-loop: Redesign processes so AI recommends, humans decide. Document the human review process.
5. Transparency notices: Update candidate communication and employee handbooks to disclose AI use in HR processes