EU AI Act and recruitment: what changes for small hiring teams

Why recruitment is classified as high risk

The AI Act sorts uses of AI by the risk they pose to people's rights. Systems used for recruitment, selection, and the evaluation of candidates sit in the high-risk category, alongside things like access to education and essential services.

The logic is straightforward: a screening error does not just annoy someone, it removes access to work, and the person affected usually never learns it happened.

What is required of you as a deployer

Most of the heavy obligations fall on whoever provides the system. As the company using it, your responsibilities are narrower but real.

• •Human oversight: a person with the authority and the information to overrule the system's output.
• •Transparency: candidates should know an AI system is being used to evaluate their application.
• •Record keeping: keep the logs the system generates, so an evaluation can be reconstructed later.
• •Use it as intended: applying a screening tool to something it was not designed for moves responsibility onto you.
• •Input data quality: if you feed it criteria that are irrelevant or discriminatory, that is your decision, not the model's.

How this overlaps with the GDPR

The two regimes reinforce each other. The GDPR already gives candidates rights over their data and, under Article 22, protection against purely automated decisions with significant effects on them. Recruitment decisions qualify.

In practice, one design choice satisfies both: keep a person responsible for the decision and keep a record of how the evaluation was produced. Everything else is documentation.

Questions to ask any AI recruitment vendor

These five questions separate tools built for the European market from tools that will hand you their compliance problem.

• •Can I see the reasoning behind an individual score, months after the fact?
• •Is there a human confirmation step before a candidate is rejected, and is it enforced?
• •Where is candidate data stored and processed?
• •What exactly is logged, and can I export the log?
• •What happens to candidate data at the end of the retention period?

How Parthumanally is built for this

Every AI evaluation records what was assessed and against which criteria, and stays attached to the application. The shortlist requires human confirmation before it drives any decision, and the reasoning is written in language a candidate could be shown.

This is not an add-on module. It is the reason the product exists in the form it does, because a screening tool for Europe cannot be retrofitted with oversight after the fact.