In short:
- The EU AI Act classifies uses according to four levels of risk.
- Tools using AI in recruitment, such as CV sorting, are high-risk systems.
- Emotion recognition in the workplace has been prohibited since 2 February 2025.
- The obligations for high-risk systems apply from 2 December 2027, according to the European Commission’s timeline.
AI in recruitment: a high-risk use
The European regulation on artificial intelligence is based on a risk approach. The European Commission distinguishes four levels.
| Risk level | Consequence |
|---|---|
| Unacceptable risk | Prohibited practices |
| High risk | Strict obligations before being placed on the market |
| Transparency risk | Disclosure obligations |
| Minimal or no risk | No specific rule |
AI in recruitment falls under high risk. The Commission expressly mentions AI tools for employment, management of workers and access to self-employment, with the example of CV-sorting software.
Prohibited practices in the workplace
The banned practices include emotion recognition in workplaces and education institutions. The prohibitions have applied since 2 February 2025.
Obligations for high-risk systems
High-risk systems must in particular meet the following requirements:
- risk assessment and mitigation;
- quality of datasets;
- detailed documentation;
- appropriate human oversight;
- a high level of robustness and accuracy.
The application timeline
| Date | Step |
|---|---|
| 2 February 2025 | Prohibited practices apply |
| 2 August 2025 | Governance rules and general-purpose AI models |
| 2 August 2026 | General application |
| 2 December 2027 | High-risk systems |
| 2 August 2028 | Systems embedded in regulated products |
This timeline is the one published by the European Commission on the date of consultation. Application is gradual: these deadlines make it possible to plan ahead for the compliance of tools the HR department already uses.
What this means for an HR department
- List the tools that use AI in the recruitment process.
- Check that the data collected remains limited to the information an employer can ask a job candidate.
- Check the compliance of HR software with personal data rules.
- Keep human involvement in decisions, including when applying job posting rules and sorting applications.
Transparency and oversight: other points to know
Transparency risk
The intermediate level of the regulation covers situations where users must know they are dealing with AI. The European Commission gives the example of chatbots: people must be informed that they are interacting with a machine.
In an HR department, a conversational agent that answers questions from candidates or employees falls under this information principle.
Governance and penalties
The regulation provides for fines in the event of non-compliance. According to the Commission, three bodies guide and advise its implementation: the AI Board, the scientific panel and the advisory forum.
What remains outside the obligations
Minimal or no-risk systems are not subject to any specific rule. Classification depends on how the tool is used, not on the technology itself.
Frequently asked questions
Is AI allowed in recruitment?
Yes, but it is regulated. The EU AI Act classifies as high-risk the AI tools used for employment, management of workers and access to self-employment, for example CV-sorting software. These systems are subject to strict obligations before they can be placed on the market.
Is emotion recognition allowed at work?
No. According to the European Commission, emotion recognition in workplaces and education institutions is one of the practices prohibited by the AI Act. These prohibitions have applied since 2 February 2025.
When do the rules on high-risk AI systems apply?
According to the timeline published by the European Commission and consulted on 4 October 2026, the rules for high-risk systems apply from 2 December 2027.
Sources
- European Commission, Législation sur l’IA, page consulted on 4 October 2026.