AI regulation enters new phase on 2 August: what employers need to know

From 2 August 2025, new measures from the European AI regulation will apply. Specifically, these concern rules on AI models for general purposes, monitoring and the associated sanctions mechanism. These changes will also affect employers’ AI policies.

Different phases

The European AI regulation provides harmonised rules for the use of artificial intelligence and entered into force on 1 August 2024. The application occurs in different phases. The rules also impact HR policies and the workplace, both for private and public employers.

On 2 February 2025 the ban on unacceptable high-risk systems, such as emotion recognition in the workplace, came into force. Since that date, employers must also invest in the AI literacy of their employees.

Soon – on 2 August 2025 – a next phase will begin. Two additional phases will follow, on 2 August 2026 and 2 August 2027.

Obligations from 2 August 2025 

From 2 August 2025, a new set of obligations will apply, mainly revolving around transparency, monitoring and enforcement. The focus is on: 

• obligations for general purpose AI model providers. These are providers like ChatGPT
• the supervision of AI systems
• the sanctions mechanism

 For most employers, it will primarily be the supervision and sanctions that are of importance.

Supervision

Both at European and national levels, the AI regulation provides for a supervisory framework, with competent institutions responsible for ensuring that the rules are correctly applied.

At the European level, an AI bureau and an AI board are already active. By 2 August 2025, however, Belgium must also designate national authorities responsible for oversight within its own country. The coalition agreement identified the Belgian Institute for Postal Services and Telecommunications for this purpose.

Sanctions

The provisions on sanctions will also apply from 2 August. On the one hand, the regulation itself provides specific maximum fines for certain offences. These apply immediately, without the intervention of the Belgian legislator:

• failure to comply with the ban on certain AI practices: fine of up to €35 million or 7% of global annual turnover (whichever is higher);
• failure to comply with obligations for, among others, providers, importers, distributors and users of high-risk AI systems: fine of up to €15 million or 3% of annual global turnover (whichever is higher); 
• providing incorrect, incomplete or misleading information to authorities: fine of up to €7.5 million or 1% of annual global turnover (whichever is higher).

On the other hand, Belgium itself has yet to lay down national rules on sanctions and other enforcement measures.

Implications for employers from 2 August

Although it is still a waiting game for Belgian legislation that concretely elaborates on the supervisory framework and the sanctions mechanism, employers will already feel the impact of this phase of the AI regulation. This means that an employer using emotion recognition in the workplace – a prohibited AI practice – risks a maximum fine of up to 35 million euros or 7% of the annual global turnover, whichever amount is higher.

In addition, other obligations continue to apply, such as enhancing AI literacy among employees and developing an AI strategy, supported by a clear AI policy.