Employment, Workers & Self-Employment , Annex III Point 4 High-Risk AI
AI used in the recruitment and selection of natural persons, for taking decisions affecting terms of work-related relationships, for task allocation based on individual behaviour, traits or characteristics, or for monitoring and evaluating performance is high-risk under Annex III Point 4.
Scope, Definitions and Boundary with Prohibited Practices
Annex III Point 4 covers AI in the full employment lifecycle: recruitment and selection, decisions affecting terms of the working relationship (promotion, termination, compensation), task allocation based on individual behaviour or personal traits, and performance monitoring and evaluation. The scope includes workers, self-employed persons, and platform workers.
Two adjacent regulations tighten the picture. Article 5(1)(f) prohibits emotion recognition in workplace contexts, so productivity tools that infer emotional state are prohibited, not merely high-risk. The Platform Workers Directive (Directive (EU) 2024/2831) imposes complementary transparency and human-oversight obligations on digital labour platforms that overlap substantially with Articles 26 and 14. Dual compliance is necessary.
Example Systems Under Annex III Point 4
Illustrative examples, each Member State's market surveillance authority may refine classification guidance over time.
Example
CV-screening AI that ranks candidates
Point 4(a) high-risk, used in recruitment and selection.
Example
Promotion-recommendation algorithm at a large employer
Point 4(b) high-risk, decision affecting terms of employment.
Example
Delivery-route assignment AI at a platform company
Point 4(c) high-risk, task allocation based on individual behaviour and traits.
Example
Keystroke-and-mouse productivity monitoring
Point 4(d) high-risk if used for performance evaluation. Emotion-inferring variants are prohibited under Article 5.
Example
AI-generated interview questions (tool only)
Not automatically high-risk, becomes high-risk if outputs materially drive selection decisions.
The Articles That Apply
High-risk classification under Annex III triggers the full Chapter III, Section 2 obligation set. Deployers pick up additional obligations under Chapter III, Section 4.
| Article | Obligation | What It Means in Practice |
|---|---|---|
| Art. 9 | Risk Management System | Particular attention to protected-attribute bias (gender, age, ethnicity, disability). Employment AI has the most extensive prior art on demographic bias, ignoring it is not defensible. |
| Art. 10 | Data Governance | Training data from past hiring decisions frequently encodes historical discrimination. Article 10(5) specifically permits processing of special category data for bias detection, but only under strict safeguards. |
| Art. 13-14 | Transparency & Human Oversight | Candidates and workers have rights under Article 26(11) to explanation of individual decisions. Oversight cannot be performative. |
| Art. 26(7) | Worker Information | Before deployment of a high-risk AI system in the workplace, workers and their representatives must be informed. This is the most frequently missed obligation. |
| Art. 26(11) & Art. 86 | Right to Explanation | Affected persons have a right to clear and meaningful explanation of individual decision-making. Black-box ranking algorithms must still deliver an intelligible account of the decision. |
| Art. 27 | FRIA | Required where the deployer is a public authority, a private entity providing a public service, or any deployer of high-risk AI systems referred to in Annex III points 5(b) and 5(c). Private employers outside these categories are not mandatorily covered by Article 27 but are strongly encouraged to conduct one. |
| Art. 49 | EU Database Registration | Standard for all Annex III systems. |
| Platform Workers Directive | Sectoral Overlap | If the deployer is a digital labour platform, Directive (EU) 2024/2831 imposes additional duties including the right to human review of significant decisions. |
Pitfalls to Avoid on Annex III Point 4
Recurrent patterns seen in pre-enforcement readiness assessments.
Emotion recognition in recruitment interviews
Inferring candidate confidence, honesty, or engagement from video is prohibited under Article 5(1)(f) when conducted in a workplace context. The workplace carve-out is interpreted broadly.
Passive monitoring dashboards claiming no 'evaluation'
If outputs are used to make decisions about workers, even indirectly, Point 4(d) applies. 'Just a dashboard' is not a defence.
Failing to inform worker representatives
Article 26(7) requires information to workers AND their representatives before deployment. Works councils that learn after the fact may obtain injunctive relief under national labour law.
Over-reliance on 'human-in-the-loop' rituals
A human who rubber-stamps AI output without independent review does not satisfy Article 14. Regulators have made this point explicitly in pre-enforcement guidance.
Operationalise Point 4 Compliance
The Full Readiness Bundle gives your legal and compliance teams the 58-Point Compliance Checklist, Annex III Classification Matrix, FRIA template, Annex IV Technical Documentation Checklist, and 7 more documents, all aligned to the Annex III high-risk regime (now applying from 2 December 2027 under the May 2026 Omnibus deal).
one-time · instant download
Get the Full Bundle →Also available: Checklist Pack $149 · White-Label $999/yr
Next Steps
Deadline Planning
15-week compliance work plan for the EU AI Act's phased timeline.
Week-by-Week Plan →