AI in hiring and HR: the rules are stricter here

The short answer

Employment is one of the areas the AI Act treats as high-risk. Systems intended for recruitment or selection, in particular for targeting job advertisements, analysing and filtering applications and evaluating candidates, and systems used for decisions on promotion and termination, task allocation based on behaviour or personal traits, and monitoring and evaluating performance, all fall within the high-risk category. For a deployer that brings obligations: use the system according to its instructions, assign human oversight to people with the competence, training and authority to exercise it, monitor operation and suspend use if a risk appears, keep the logs the system generates, and inform workers and their representatives before putting such a system into use. Data protection adds a second layer, including a basis, an impact assessment in most cases, transparency to candidates, and the restrictions on solely automated decisions producing significant effects. Emotion recognition in the workplace is prohibited with narrow exceptions. For a small business the sensible line is to keep AI out of selection and use it for the administration around hiring instead.

What is high-risk and what is not

Use	Classification	Position for a small business
Filtering, ranking or scoring applicants	High-risk	Avoid; or treat as a full compliance project
Targeting job advertisements	High-risk	Care with platform tools that do this automatically
Automated assessment or testing of candidates	High-risk	Avoid unless properly governed
Promotion, termination and task allocation decisions	High-risk	Human decisions with documented criteria
Performance monitoring and evaluation	High-risk	Legal advice and consultation first
Emotion recognition in the workplace	Prohibited, with narrow exceptions	Do not
Drafting job descriptions and adverts	Minimal	Useful; review for biased language
Scheduling interviews and communication	Minimal	Useful; saves real time
Transcription and structured interview notes	Minimal, with data protection care	Useful; interviewers assess
Summarising a candidate’s own submitted material for the panel	Borderline; keep supportive	Panel reads the source
Onboarding administration	Minimal	Useful

Where AI helps safely in hiring

Our position

We advise clients to keep AI out of candidate selection. The obligations are substantial, the bias risk is real and difficult to test at small volumes, the consequences of getting it wrong reach individuals’ livelihoods, and the efficiency gain over structured human screening is modest for the number of applications a small business receives. The administration around hiring, on the other hand, consumes real hours and carries almost no risk, which is where we help clients apply it.

What this means for you

Employment uses of AI, screening, selection, promotion, termination, task allocation and monitoring, are high-risk, with deployer obligations including instructions, real human oversight, monitoring, logs and informing workers, alongside data protection requirements. Emotion recognition at work is prohibited with narrow exceptions. Keep AI out of selection, use it for the administration around hiring, and check what your HR software’s AI features are already doing. This is general information rather than legal advice; employment AI warrants qualified input.