EU AI Act Brief – Part 4: AI at Work
The EU AI Act represents a landmark regulation aimed at addressing the deployment of AI systems in various contexts, including the workplace. This brief focuses on the implications of the AI Act for workers, particularly in terms of protections and obligations for employers.
Introduction
In recent years, the use of algorithmic management and decision-making systems has surged in workplaces. A recent OECD survey revealed that over 70% of managers reported utilizing at least one automated tool to instruct, monitor, or evaluate employees. This trend has raised significant concerns regarding the rights and protections of workers.
Worker Sentiment and the EU AI Act
A survey conducted by the European Commission highlighted overwhelming support among workers for regulations governing the use of AI in the workplace, particularly calls for a Directive addressing the challenges posed by algorithmic systems.
The EU AI Act recognizes the risks associated with AI systems in employment and introduces specific obligations aimed at safeguarding workers. This comprehensive approach includes prohibitions on certain high-risk AI practices.
Prohibited AI Systems
The AI Act adopts a risk-based approach, prohibiting several AI practices considered to pose unacceptable risks in the workplace. Key prohibitions include:
1. Biometric Categorization
The Act entirely prohibits AI systems that categorize individuals based on their biometric data to deduce attributes such as race or trade union membership. This prohibition safeguards workers from negative consequences related to their identification.
2. Emotion Recognition
Emotion recognition systems are mostly prohibited in employment settings. The Act acknowledges the unreliability of these systems and permits their use only for medical or safety reasons. Exceptions must be strictly necessary and accompanied by adequate safeguards.
3. Social Scoring
The AI Act prohibits social scoring systems that lead to detrimental treatment based on social behavior. This is particularly relevant in employment, where unfair treatment based on inferred personality traits can occur.
4. Real-Time Biometric Identification
While the Act prohibits real-time remote biometric identification in law enforcement, it permits its use in other contexts, potentially allowing employers to surveil workers.
Limited Protections from High-Risk Systems
The majority of the AI Act focuses on regulating high-risk AI systems, which are allowed but subject to strict safeguards. An AI system is classified as high-risk if it falls under specific categories outlined in the Act, including recruitment and performance evaluation.
Employers deploying high-risk AI systems are bound by obligations that enhance transparency and accountability. These include ensuring human oversight and providing notice to affected workers before implementation.
Obligations for Deployers
General Obligations
The AI Act imposes general obligations on all deployers, including:
- Ensuring human oversight of AI systems.
- Informing workers’ representatives and affected workers about the deployment of AI systems.
Obligations for Public Authorities
Deployers who are public authorities face heightened obligations, including conducting a fundamental rights impact assessment (FRIA) prior to deploying high-risk AI systems. This requirement extends to private entities performing public functions.
Remedies and Enforcement
The AI Act offers limited remedies for individuals subjected to high-risk AI systems, such as the right to an explanation regarding decisions made based on AI outputs. Individuals can also lodge complaints with national authorities if they believe the Act has been violated.
Oversight mechanisms, including the designation of fundamental rights authorities, facilitate the enforcement of the AI Act, ensuring that workers’ rights are protected in the context of AI deployment.
Conclusion
While the EU AI Act marks a significant advancement in regulating AI in the workplace, its effectiveness in protecting workers’ rights is limited. The Act introduces essential restrictions on AI practices, notably in emotional recognition, and mandates notice and explanation requirements for workers. However, significant loopholes exist, particularly regarding the obligations of private employers. Ongoing monitoring of the implementation of the AI Act and its impact on workers is crucial for ensuring that their rights are adequately safeguarded in the evolving landscape of AI technology.
