AI Regulation: Sanctions in force since August 2, 2025

The European AI Regulation (AI Act), adopted in June 2024, will enter into force in phases. Following the initial obligations that have been in effect since February 1, 2025, a new phase will begin on August 2, 2025 , when the provisions on sanctions will take effect.

Background

The AI Regulation formally entered into force in the European Union on August 1, 2024. The aim of this legislation is to create a uniform legal framework for the use of artificial intelligence. The EU aims to leverage the benefits of AI while simultaneously mitigating its risks and promoting safe and ethical use.

The first obligations for companies, which have been in force since February 2025, concern:

• The ban on AI systems that pose an unacceptable risk (e.g. social scoring).

• The training obligation for employers to make employees AI-literate.

  
  

New phase from August 2, 2025

From August 2025, new provisions will be added that relate to, among other things:

• Generic AI models (foundation models).

• Notifying authorities and notified bodies , which assess high-risk systems.

• Governance and confidentiality .

• And most importantly: sanctions (except those specifically for generic AI models).

For employers and organisations, the sanction provisions are the most concrete and urgent.

The AI Act sanctions regime

The AI Regulation obliges all EU Member States to establish national rules on sanctions and enforcement. This includes not only fines, but also warnings and other corrective measures.

The sanctions must be effective, proportionate, and dissuasive . At the same time, the financial capacity of SMEs and startups must be taken into account. In Belgium, specific national measures are still being developed, but the regulation itself already contains clear maximum amounts.

Maximum administrative fines

• Prohibited AI applications (unacceptable risk): Up to €35,000,000 or 7% of global turnover (whichever is higher, with exceptions for SMEs/start-ups).

• Breach of obligations in high-risk systems: Up to €15,000,000 or 3% of global turnover.

• Incorrect or misleading information to regulators: Up to €7,500,000 or 1% of global turnover.

  
  

The final sanction depends on the context. Factors such as the severity and duration of the violation, the number of people affected, the extent of the damage, and whether the violation was intentional or negligent all play a role.

Important to remember

Since August 2, 2025, companies have for the first time faced administrative fines for violations of the AI Regulation. This provides regulators with a powerful tool to enforce compliance.

The Belgian legislature is still working on the national implementation of this European sanctions mechanism. Nevertheless, it is already crucial for organizations to review their AI systems and processes and be prepared for this next phase.

Would you like to stay informed about everything related to social security legislation or payroll? Contact Rovoco!