These days, Artificial Intelligence (AI) is everywhere. In fact, I used an AI tool myself to help organise my thoughts for this article. AI has also become an indispensable part of the modern workplace, particularly in HR departments. However, Belgian labor law, influenced heavily by the new European AI Act, imposes specific obligations and risks that HR must consider before integrating AI tools into the daily operations.
The European AI Act: core principles and risk levels
The European AI Act, which entered into force in Belgium on February 2, 2025, sets the framework for AI use at work. It classifies AI systems into four risk levels:
- Minimal Risk: Most general AI systems, such as basic chatbots and spam filters, fall here and face minimal regulation;
- Limited Risk: AI systems that interact directly with humans (e.g. chatbots) must inform users that they are engaging with AI;
- High Risk: Systems used in critical HR processes, like CV screening or automated candidate evaluation, are considered high-risk, requiring robust documentation, human oversight, and transparency. Strict requirements for data quality and user information apply. These requirements will be enforced even more strictly from August 2026 onwards;
- Unacceptable Risk: AI tools infringing on fundamental rights—such as systems for emotion recognition or social scoring—are outright banned on the workplace.
Key legal obligations for Belgian employers
1. AI literacy for employees (as of February 2025)
Employers must ensure AI literacy among all staff (including permanent, freelancers, and temporary workers) who use AI systems. The required knowledge level varies depending on the user’s interaction with AI. Those directly involved in decision-making processes with AI need a deeper understanding than occasional users. Training and open communication are essential to fulfill this obligation.
2. Transparency and internal policies
HR managers are encouraged to adopt a transparent approach. Employees must be explicitly informed about when and how AI is being employed – especially in sensitive areas such as recruitment, evaluations, or workplace surveillance. It is advisable to implement an AI policy clarifying permitted and prohibited uses, monitoring, and a contact person for questions or concerns.
3. Employee involvement and social dialogue
The Belgian legal landscape emphasizes employee engagement, especially when introducing significant workplace changes like AI adoption.
In this regard, we rely on existing regulations. The adoption of new technologies in the workplace is governed by the national collective bargaining agreement (cba) n°39. This cba is specifically drafted to ensure that employees are properly informed and consulted about significant technological innovations, that could affect their working conditions or job security. Although the cba was implemented long before AI made his way to the work floor, its provisions are still valid today. When an employer plans to implement new digital tools or AI applications with significant impact on the employment, he is legally obliged to initiate an information and consultation procedure with the works council or trade union delegation (if any) and this prior to the implementation of the new technology.
4. Monitoring and compliance
Employers must consistently monitor the deployment of AI systems, especially those in the high-risk category. Prohibited uses (e.g., tools that analyze emotional states or attempt to predict future behaviors based on social scoring) must be identified and immediately banned. Violations can result in severe penalties starting August 2026, with fines up to €35 million in certain cases.
5. Respecting the GDPR: the right to human intervention
Crucially, the use of AI in the workplace under both Belgian law and the new European AI Act must always respect the existing guarantees in the General Data Protection Regulation (GDPR). The GDPR explicitly gives individuals the right not to be subject solely to automated decisions -including those based on AI – that have legal effects or significant consequences for them, such as hiring decisions or performance evaluations. Employees always have the right to request meaningful human intervention if they disagree with an automated decision, to express their point of view, and to contest the outcome.
Best practices for HR
- Start with a clear purpose: define how AI can improve employee experience and business objectives;
- Develop and update your AI policy: ensure ongoing compliance with EU and national regulations;
- Invest in training: focus on practical, role-specific AI literacy for all affected employees;
- Foster an open, transparent culture: keep communication clear regarding AI’s use, benefits, and risks;
- Engage employees early: use workshops, surveys, and dialogue to minimize resistance and gather feedback.
Conclusion
AI presents both opportunities and legal complexities for the HR department in Belgium. Proactive adaptation, transparency, and a commitment to staff development are essential to ensuring a compliant, ethical, and productive AI-driven workplace.
Yne Machiels – Attorney-Partner
www.reliancelaw.be