What is EU AI Act Annex III point 4?
Annex III point 4 of the EU AI Act lists employment and workers management as a high-risk use area. It covers AI systems intended for recruitment and selection (point 4(a)) and AI systems intended to make work-relationship decisions, allocate tasks based on individual behaviour or traits, or monitor and evaluate performance and behaviour (point 4(b)). A listed system is high-risk unless the Article 6(3) exemption applies. Verify with counsel.
- 2 August 2026 — the date from which Annex III high-risk system obligations apply to providers and deployers of employment and workers management AI under Article 6(2); prohibited-practice rules under Article 5 — including emotion-recognition AI in the workplace — apply from 2 February 2025; verify the applicable phase-in date for your system class with counsel (EU AI Act 2024/1689, OJ L 2024/1689, 12 July 2024).
- EUR 15 million or 3% of worldwide annual turnover — maximum administrative fine for violations of most Annex III high-risk system obligations under Article 99(4), whichever is higher; EUR 35 million or 7% of worldwide turnover for Article 5 prohibited-practice violations under Article 99(3); the turnover-linked ceiling means large employers carry the largest exposure (EU AI Act 2024/1689; verify applicable ceiling with counsel).
- Eight Annex III use areas — the EU AI Act Annex III lists eight high-risk use categories; point 4 covers employment and workers management specifically; the other seven span biometric identification, critical infrastructure management, education and vocational training, essential private and public services, law enforcement, migration and asylum, and administration of justice; point 4 is the category directly engaged by workplace productivity monitoring (EU AI Act 2024/1689, Annex III; verify classification of your specific system with counsel).
Which workplace AI systems are high-risk under Annex III point 4?
Annex III point 4 lists two categories. Point 4(a) covers recruitment and selection systems — targeted job advertising, application filtering, and candidate evaluation. Point 4(b) covers systems that make decisions on terms, promotion, or termination of work relationships, allocate tasks based on behaviour or traits, or monitor and evaluate performance and behaviour. Workplace productivity monitoring sits in point 4(b).
Mapping common workplace systems onto the two sub-categories clarifies the boundary:
| Workplace system | Annex III point 4 sub-category | Listing position |
|---|---|---|
| AI CV screening / application filtering | 4(a) recruitment and selection | Listed |
| Targeted job-advert placement engine | 4(a) recruitment and selection | Listed |
| AI performance scoring / productivity evaluation | 4(b) monitor and evaluate performance and behaviour | Listed |
| Task-allocation engine based on individual behaviour | 4(b) task allocation on behaviour or traits | Listed |
| Promotion / termination decision-support AI | 4(b) work-relationship decisions | Listed |
| Descriptive dashboard with no evaluation or decision role | Not engaged by the listing on its face | Assess against 4(b) and Article 6(3) |
The listing turns on the system's intended purpose, not its marketing label. A tool sold as a "dashboard" that in fact scores or ranks workers for evaluation is engaging point 4(b). For the formal classification logic, the companion guide on Article 6 high-risk classification for workforce AI walks the full Article 6 route.
Is productivity monitoring high-risk under the EU AI Act?
AI systems used to monitor and evaluate the performance and behaviour of persons in a work relationship are listed high-risk under Annex III point 4(b). Productivity monitoring that uses AI to evaluate performance therefore falls within the high-risk category, subject to the Article 6(3) exemption. The provider carries the full obligations stack; the deployer carries derivative Article 26 obligations.
The decisive word is evaluate. A system that captures activity and renders it descriptively, without AI-driven evaluation of the person's performance, has a weaker claim to be inside 4(b) on its face — but the moment the system scores, ranks, or infers performance, it is squarely listed. The design lesson many EU-facing employers draw is to keep evaluation human-led and reserve the AI layer for surfacing behaviour signals rather than rendering the verdict, which is also where the Article 6(3) exemption analysis becomes relevant.
What is the Article 6(3) exemption to Annex III high-risk classification?
Article 6(3) allows an Annex III-listed system to be treated as not high-risk if it does not pose a significant risk of harm to health, safety, or fundamental rights, including by not materially influencing decision-making. The conditions: a narrow procedural task, improving a completed human activity, detecting patterns without replacing or influencing the human assessment, or a preparatory task. A profiling system is always high-risk and cannot use the exemption.
The four exemption conditions, applied to workplace systems:
- Narrow procedural task. The system performs a defined, mechanical step — for example, de-duplicating records — rather than evaluating the person.
- Improving a completed human activity. The system refines an output a human has already produced, without substituting its own judgement for the human's.
- Pattern detection without influence. The system surfaces patterns for human consideration but does not replace or materially influence the human assessment.
- Preparatory task. The system prepares inputs for a decision a human makes, without making or steering the decision.
Assuming a human sign-off unlocks the exemption. Article 6(3) is not satisfied by adding a human approver. If the system scores, ranks, or recommends in a way the human relies on, it materially influences the outcome and stays high-risk. And any system that performs profiling of natural persons is always high-risk — the exemption is closed to it regardless of human involvement. Human oversight under Article 14 is a separate obligation, not an exemption.
How should an India IT exporter classify a workforce AI system under Annex III?
India IT exporters classify in three gates. First, the Annex III gate — does the intended purpose fall within point 4(a) or 4(b)? Second, the Article 6(3) exemption gate — does it perform only a narrow procedural, preparatory, or pattern-detection task without materially influencing the decision, and avoid profiling? Third, the role gate — is the firm provider, deployer, or both? The classification determines the obligations stack.
India IT services firms touching EU-domiciled workers fall within the Act's extraterritorial scope under Article 2. The classification outcome drives which obligations attach:
| Role | Core obligation if system is high-risk | Key articles |
|---|---|---|
| Provider (builds / supplies the system) | Full high-risk stack: risk management, data governance, transparency, conformity assessment, registration | Articles 9, 10, 13, 16, 43 |
| Deployer (operates the system on EU workers) | Use per instructions, human oversight, monitoring, records, worker information | Articles 26, 14 |
| Both provider and deployer | Both stacks, with the FRIA where Article 27 scopes it | Articles 9, 26, 27 |
Document the three-gate assessment as a registered artefact. The EU AI Act Article 6 Classifier runs the Annex III plus Article 6(3) logic; the EU AI Act Vendor Scorecard screens the provider posture once classification lands. Verify with counsel.
Run the Annex III classification and the vendor screen
Classify the system against Annex III point 4 and Article 6(3), then score the provider. Both free; email-gated only at PDF download.
Frequently asked questions
Which workplace AI systems are high-risk under EU AI Act Annex III point 4?
Annex III point 4 of the EU AI Act lists two categories of employment and workers management AI as high-risk. First, AI systems intended to be used for the recruitment or selection of natural persons, in particular to place targeted job advertisements, analyse and filter job applications, and evaluate candidates. Second, AI systems intended to be used to make decisions affecting terms of work-related relationships, the promotion or termination of work-related contractual relationships, to allocate tasks based on individual behaviour or personal traits, or to monitor and evaluate the performance and behaviour of persons in such relationships. Workplace performance monitoring and productivity scoring fall in the second category. Verify with counsel.
Is productivity monitoring high-risk under the EU AI Act?
AI systems used to monitor and evaluate the performance and behaviour of persons in a work-related relationship are listed high-risk under Annex III point 4(b). Productivity monitoring that uses AI to evaluate performance therefore falls within the high-risk category, subject to the Article 6(3) exemption assessment. The provider carries the full high-risk obligations stack including risk management, data governance, transparency, and human oversight; the deployer carries derivative Article 26 obligations. A purely descriptive dashboard that does not evaluate or make decisions may sit outside the listing, but the assessment is fact-specific. Verify with counsel.
What is the Article 6(3) exemption to Annex III high-risk classification?
Article 6(3) allows an AI system listed in Annex III to be treated as not high-risk if it does not pose a significant risk of harm to health, safety, or fundamental rights, including by not materially influencing the outcome of decision-making. The Act gives conditions: the system performs a narrow procedural task, improves the result of a previously completed human activity, detects decision-making patterns without replacing or influencing the human assessment, or performs a preparatory task. Crucially, a system that performs profiling of natural persons is always high-risk and cannot use the exemption. The provider documents the exemption assessment and registers it. Verify with counsel.
Does Annex III point 4 apply to India IT services firms?
India IT services firms that provide or deploy workplace AI systems touching EU-domiciled workers fall within the EU AI Act's extraterritorial scope under Article 2. A firm that builds or supplies a workforce AI system used in the EU is a provider with Annex III high-risk obligations; a firm that operates such a system on EU workers as part of a customer engagement is a deployer with Article 26 obligations. India IT exporters classifying systems for EU customers run the Annex III point 4 test plus the Article 6(3) exemption assessment to determine the obligations stack. Verify with counsel.
What is the difference between Annex III listing and high-risk classification?
Annex III listing is the first gate — the AI system's intended purpose falls within one of the listed use cases, such as employment and workers management under point 4. High-risk classification is the outcome after the Article 6(3) exemption assessment — a listed system is high-risk unless it qualifies for the exemption by not posing a significant risk and not profiling. So Annex III listing triggers the assessment; classification is the result. A system can be Annex III-listed but classified not high-risk if it clears Article 6(3), provided it does not profile.
When does Annex III high-risk classification start to be enforced?
The EU AI Act applies the high-risk obligations for Annex III systems from 2 August 2026, with the broader framework phasing in across 2025 and 2026. Providers and deployers of workplace AI systems classified high-risk under Annex III point 4 build to the obligations stack ahead of that window. Specific transitional provisions apply to systems already on the market, so confirm the applicable date for a given system with counsel.
How should an India IT exporter classify a workforce AI system under Annex III?
Run the classification in three gates. First, the Annex III gate — does the system's intended purpose fall within point 4(a) recruitment or point 4(b) decisions, task allocation, or performance monitoring? Second, the Article 6(3) exemption gate — does it perform only a narrow procedural, preparatory, or pattern-detection task without materially influencing the human decision, and does it avoid profiling? Third, the role gate — is the firm the provider, the deployer, or both? The classification determines the obligations stack. Document the assessment as a registered artefact. Verify with counsel.
Can a workplace AI system avoid high-risk classification by keeping a human in the loop?
Human involvement alone does not defeat high-risk classification. The Article 6(3) exemption requires that the system not materially influence the outcome of decision-making, for example by detecting patterns without replacing or influencing the human assessment. A system that scores, ranks, or recommends in a way the human decision-maker relies on materially influences the outcome and remains high-risk even with a human signing off. And a profiling system is always high-risk regardless. Human oversight under Article 14 is a separate obligation, not an exemption route. Verify with counsel.
Disclaimer. This deep dive reflects the EU AI Act 2024/1689 as published in the Official Journal of the European Union; secondary legislation (delegated acts, implementing acts, and Commission guidelines on Article 6(3)) is expected and may refine the classification specifics. Annex III listing and the Article 6(3) exemption assessment are fact-specific to the system's intended purpose. Verify all classification outcomes with EU counsel for EU deployment and with India counsel for India IT services exporter engagements before relying on any output in a regulatory submission. Questions: hello@gstride.ai.