Knowledge base
AI in HR: what can an SME do and what does the AI Act ask?
By Bob Biesemans · published on
Writing a job ad with AI is fine. Having AI rank candidates falls under stricter rules. What the AI Act asks of employers after the Digital Omnibus, and what you can sort out now.
Short answer
Using AI for everyday HR writing, such as a job ad, is fine within your company's rules. AI that filters CVs, ranks candidates or monitors employee performance is high-risk AI; under the Digital Omnibus those obligations apply from 2 December 2027. Emotion recognition in the workplace has been prohibited since 2 February 2025.
Not every use of AI in HR is equally sensitive
HR staff often use AI for everyday writing: drafting a job ad, rewriting an onboarding booklet, translating a job description. The same rules apply as in the rest of the company: no personal data in tools that aren't approved for it, and read the result before it goes out.
It changes as soon as AI helps decide about people. The AI Act classes AI systems that filter CVs, rank candidates, decide on promotion or dismissal, allocate tasks based on behaviour or monitor employee performance as high-risk (Annex III, point 4).
What is already prohibited
Since 2 February 2025 it has been prohibited to use AI to infer employees' emotions in the workplace, except for medical or safety reasons (Article 5(1)(f)). A tool that reads employees' mood from their voice, face or typing is therefore off limits. The same goes for job interviews in which software would assess candidates' emotions.
What comes on top from 2 December 2027
Under the Digital Omnibus (Regulation (EU) 2026/1744), the obligations for Annex III high-risk AI apply only from 2 December 2027, instead of August 2026. Postponed is not cancelled: anyone using such a system will then have to use it according to the provider's instructions, assign human oversight to people with the necessary training and authority, and inform employees and their representatives beforehand (Article 26).
The GDPR also still applies. A decision with legal effects, such as a rejection, may not be based solely on automated processing without the safeguards of Article 22 GDPR. And in Belgium, introducing new technology at work can also require informing and consulting employee representatives.
What you can do as an SME now
List the AI features in your HR software and recruitment channels; they are often built in without anyone switching them on deliberately. Agree which HR tasks AI may support and where a person always decides. Ask suppliers which AI features their product has and how to switch them off or check them.
Above all, make sure the people using these tools know where the limits are: an AI ranking is a suggestion, not a judgement. That is exactly what Article 4 of the AI Act asks: measures for the AI literacy of those who use AI. Measure it with tasks from HR work itself, explain where needed and record what you did.
Frequently asked questions
Can I use ChatGPT to write a job ad?
Yes. It is an everyday writing task. Don't put personal data into it unless the tool is approved for that, and check the result for errors and for wording that would exclude certain candidates.
Is AI CV screening prohibited?
No, but it is high-risk AI (Annex III of the AI Act). From 2 December 2027 you must use such a system according to its instructions, provide human oversight and inform employees beforehand. The GDPR already prohibits a rejection based solely on automated processing without safeguards.
What should I do now?
Map the AI features in your HR tools, agree where a person always decides, and make sure those using the tools know the limits. That last part is what Article 4 asks: measures for AI literacy, which you record.