The first substantive provisions of the AI Act entered into force on February 2, 2025 (see our post EU AI Act Published: Dates for Action for a more detailed timeline). The AI Act adopts a risk-based approach, categorizing AI systems based on their associated risks, and several employment-related use cases are considered as higher risk or prohibited entirely. Consequently, employers should authorize the use of AI systems in the workplace only after a thorough legal review.
Here is a summary of the key considerations for employers when implementing AI systems in the workplace:
Obligations already in force
- Employers must ensure that their staff possess the necessary knowledge and skills to operate the AI systems used by the company. This AI competence can be developed through information sessions and regular training courses. Employers must also provide employees with the operating instructions for any AI system.
- Certain AI systems, such as those designed to recognize employees’ emotions, are prohibited in the workplace from the outset.
Obligations you should be preparing for now
- AI systems used in the HR sector are frequently deemed high-risk. High-risk AI systems include those used for assessing, selecting, or recruiting applicants, as well as systems making decisions related to promotions, dismissals, or working conditions.
- High-risk AI systems require human oversight. The supervisor must have the necessary skills and training and be authorized to make any necessary corrections when using the system.
- Data entered into the AI system by companies must fulfill the system’s purpose and be representative. Data is considered representative if all groups for whom the AI system is intended are adequately reflected statistically.
- Prior to deploying a high-risk AI system, it is essential to inform the works council, if any, and the affected employees about its use. Furthermore, if the AI system is capable of monitoring or processes personal employee data, a works council agreement must be established.
Recommendation
We advise companies to proactively establish a strong compliance system for their AI systems. Without such a system, the benefits of AI may be outweighed by the significant risk of penalties under the AI Act, which can impose fines of up to EUR 35 million or 7% percent of annual global turnover for violations.
Click here to read the German version.
