The new AI regulation - what applies now?

Why Is an AI Regulation Needed?

In recent years, the use of AI across various parts of society has increased rapidly. In recruitment, AI has become a tool to handle large volumes of candidates, reduce human errors, and streamline selection processes. But alongside the technology, the risks have also become apparent.

Studies have shown that certain AI systems used in recruitment may instead reinforce existing biases – for example, by systematically favoring gender, ethnicity, or age. Algorithms trained on historical data from organizations with unequal representation have been able to reproduce the same distortions in future selections. Several high-profile cases where AI-driven systems have discriminated against candidates have raised concerns among both the public and lawmakers.

It is against this backdrop that the AI Act has been developed. The EU’s goal is to ensure that the technology is transparent, fair, safe, and respectful of individual rights.

AI in Recruitment = High Risk According to the AI Act

The AI Act is based on a risk-based approach where AI systems are categorized into four risk levels: unacceptable, high, limited, and minimal risk. Systems used to make decisions in recruitment processes (e.g., screening, scoring, or analyzing candidate behavior) fall under high risk. This means their use is regulated with strict requirements for documentation, monitoring, transparency, and fairness.

Using AI in recruitment is not prohibited, but it now comes with higher demands. For example:

• Automated selection systems must be transparent – candidates should understand that AI is being used and be able to request insight into how decisions are made.
• Human oversight is mandatory – decisions that affect people cannot be made without the ability for a human to review and challenge them.
• Data quality and non-discrimination must be ensured – AI may not be trained on biased data sets that could lead to skewed outcomes.
• Everything must be documented – from training data to decision logic and risk assessments.

In short, AI cannot be used as a black box. It must be explainable, fair, and controllable.

What Is Allowed – and What Is Not?

Recruitment companies may still use AI to streamline processes, for example by:

• Screening and ranking resumes
• Analyzing candidates’ responses in tests or interviews
• Improving candidate-role matching

However, it will be prohibited or severely restricted to:

• Let AI make decisions entirely without human involvement
• Use AI models that select candidates based on sensitive variables (such as gender, ethnicity, or age) – even if these are not explicitly input
• Use manipulative systems, such as AI that interprets micro facial expressions without scientific foundation

What Does This Mean for Recruitment Companies?

Companies that use or develop AI in recruitment need to adapt their processes and tools:

1. Identify where AI is used – in which parts of the process and in which tools.
2. Map the risks – is the technology fair, transparent, and explainable?
3. Create documentation – how does the system work, what data is used, how is effectiveness evaluated?
4. Introduce human oversight – every candidate should be able to understand and challenge assessments.
5. Ensure the candidate experience is ethical and respectful – clearly indicate when AI is being used.

The AI Act is not a ban on AI in recruitment – quite the opposite. It is a guide toward responsible, ethical, and transparent use. For the recruitment industry, it marks a paradigm shift where trust, fairness, and candidates’ rights become as important as efficiency. The companies that manage to combine these perspectives will be the strongest players in the new AI era.