Key Definitions
| Term | Definition |
|---|---|
| Automated Employment Decision Tool (AEDT) | Any computational process, derived from machine learning, statistical modeling, data analytics, or artificial intelligence, that issues simplified output (score, classification, or recommendation) used to substantially assist or replace discretionary decision-making for employment decisions. (NYC Local Law 144 definition) |
| High-Risk AI System (Employment) | Under EU AI Act Annex III, point 4: AI systems intended for use in employment, workers management, and access to self-employment, including recruitment, CV screening, job advertisement targeting, interview/test analysis, promotion and termination decisions, task allocation, and performance monitoring. |
| Algorithmic Bias Audit | A systematic evaluation of an AI employment tool's outputs to identify whether the tool produces statistically significant disparities across protected demographic groups. |
| Disparate Impact | A form of discrimination that occurs when a facially neutral employment practice disproportionately disadvantages members of a protected group, regardless of intent. |
| Four-Fifths (80%) Rule | A practical guideline from the US EEOC Uniform Guidelines on Employee Selection Procedures: a selection rate for any protected group that is less than four-fifths (80%) of the selection rate for the group with the highest rate is generally regarded as evidence of adverse impact. |
| Human-in-the-Loop (HITL) | An AI deployment model where a human reviewer is actively involved in the decision-making process, reviewing AI outputs before final employment decisions are made. |
| Worker Notification | The legal requirement to inform workers, job candidates, or their representatives about the use of AI systems in employment decisions, including the nature, purpose, and implications of the AI processing. |
| Algorithmic Management | The use of AI and automated systems to direct, monitor, evaluate, and discipline workers, including task allocation, performance tracking, scheduling, and productivity measurement. |
| Fundamental Rights Impact Assessment (FRIA) | A structured evaluation required by EU AI Act Article 27 for deployers of high-risk AI systems (when applicable) to assess impacts on fundamental rights before deployment. |
| Works Council Consultation | The legal requirement in many EU Member States to consult with employee representative bodies before introducing AI systems that affect working conditions, including monitoring and performance evaluation systems. |
| Emotion Recognition System | An AI system that identifies or infers emotions, intentions, or mental states of natural persons. Under the EU AI Act, use of emotion recognition in workplace and education settings is restricted. |
| Platform Worker | A person performing platform work, regardless of the contractual relationship, whose work is organized through a digital labor platform. Subject to the Platform Workers Directive provisions on algorithmic management. |
Chapter 1: AI in Employment — The High-Risk Classification
AI systems used in employment are classified as high-risk under the EU AI Act, making them subject to the most stringent compliance requirements in the regulation. This classification covers the entire employment lifecycle — from recruitment advertising through hiring, onboarding, performance management, promotion, task allocation, and termination. Organizations deploying AI in any employment context must implement comprehensive compliance measures or face penalties of up to 15 million euros or 3% of global annual turnover.
1-1. EU AI Act Annex III, Point 4 — What Is Covered
The EU AI Act classifies the following AI employment uses as high-risk:
| Use Case | Annex III Reference | Examples |
|---|---|---|
| Recruitment and selection | Point 4(a) | CV screening tools, candidate matching algorithms, video interview analysis, psychometric assessment AI, chatbot-based screening |
| Job advertisement targeting | Point 4(a) | AI that determines which candidates see job postings, targeted recruitment algorithms |
| Decision-making on applications | Point 4(a) | AI scoring of applicants, recommendation engines for hiring managers, automated shortlisting |
| Promotion and termination decisions | Point 4(b) | AI-driven performance ratings that inform promotion decisions, predictive attrition models used for termination planning |
| Task allocation | Point 4(b) | Algorithmic task assignment in warehouses, delivery routing, gig platform work allocation |
| Performance and behavior monitoring | Point 4(b) | Productivity tracking AI, keystroke monitoring, camera-based behavior analysis, email sentiment analysis |
| Work relationship evaluation | Point 4(b) | AI assessing worker engagement, team dynamics analysis, relationship mapping |
1-2. Why Employment AI Is High-Risk
The EU AI Act classification reflects the severe potential impacts of AI in employment:
- Power asymmetry. The employer-employee relationship involves inherent power imbalance. AI amplifies this asymmetry by enabling surveillance and decision-making at scale.
- Livelihood impact. Employment decisions directly affect individuals' economic well-being, career trajectory, and quality of life.
- Discrimination risk. Historical employment data frequently encodes patterns of discrimination. AI systems trained on such data can perpetuate and amplify these patterns.
- Opacity. Many AI employment tools use complex models whose decision logic is difficult to explain to candidates and workers.
- Scale. AI enables employers to make decisions about thousands of individuals simultaneously, amplifying any systematic bias.
1-3. Timeline and Enforcement
- 2 February 2025: Article 4 (AI literacy) became enforceable — all organizations using AI in employment must ensure relevant staff understand the systems.
- 2 August 2026: Full high-risk obligations for employment AI become applicable — the complete compliance framework must be implemented.
- Enforcement: National market surveillance authorities will enforce EU AI Act compliance. Penalties for non-compliance with deployer obligations can reach 15 million euros or 3% of global annual turnover.
1-4. The Broader Regulatory Landscape
Employment AI faces regulation from multiple directions simultaneously:
| Regulation | Jurisdiction | Focus | Status |
|---|---|---|---|
| EU AI Act Annex III | EU/EEA | Comprehensive AI regulation, employment as high-risk | High-risk obligations from 2 August 2026 |
| GDPR Art.22 | EU/EEA | Automated individual decision-making | In force |
| Platform Workers Directive | EU | Algorithmic management of platform workers | Adopted 2024, transposition by 2 December 2026 |
| NYC Local Law 144 | New York City | Automated employment decision tools — bias audits | In force since July 2023 |
| EEOC AI Guidance | US (federal) | AI and Title VII anti-discrimination compliance | Guidance published May 2023 |
| Illinois AI Video Interview Act | Illinois | AI analysis of video interviews | In force since January 2020 |
| Colorado AI Act (SB 24-205) | Colorado | AI in consequential decisions including employment | Effective 1 February 2026 |
| Employment Equality Directive | EU | Non-discrimination in employment | In force |
| National works council laws | Various EU Member States | Worker consultation rights for new technology | In force (varies by country) |