The EU AI Act in HR and recruitment: what employers need to know
The AI Act is the European Union regulation on artificial intelligence (Regulation 2024/1689). It sorts AI systems by risk and sets rules accordingly. AI systems for recruiting, selecting and assessing people are classed as high-risk. Companies that use them will have to ensure human oversight, inform candidates and employees and keep records.
What is the AI Act?
The AI Act is the first comprehensive law on artificial intelligence in the European Union. As a regulation, it applies directly in all member states. It entered into force on 1 August 2024 and its rules apply in stages.
The regulation distinguishes four risk levels.
| Level | Meaning | HR example |
|---|---|---|
| Unacceptable risk | The practice is prohibited. | AI that recognizes employees’ emotions. |
| High risk | Allowed, but with strict obligations. | Screening resumes, assessing candidates, deciding on promotions. |
| Limited risk | Transparency obligation. | A chatbot or voicebot that talks to candidates. |
| Minimal risk | No specific obligations. | Spell-checking a job ad. |
Status as of 18 November 2026. This text reflects the changes made by Regulation 2026/1744. It is for information only and does not constitute legal advice.
Why does the AI Act apply to recruitment?
Annex III of the regulation lists the areas in which AI is considered high-risk. One of them is employment and workers management. It covers AI systems that:
- are used to recruit or select people, for example to place targeted job ads, analyze and filter job applications or evaluate candidates,
- influence decisions on working conditions, promotion or termination of employment,
- allocate tasks based on behavior or personal traits,
- monitor and evaluate the performance and behavior of employees.
Not every AI tool in HR is high-risk, though. Under Article 6(3), a system may not be high-risk if it does not materially influence the decision, for example if it only performs a narrow procedural or preparatory task. This exception does not apply if the system profiles people. The provider of the system makes and documents this assessment.
Timeline: what applies when
The rules for high-risk AI were originally due to apply from 2 August 2026. Regulation 2026/1744, known as the Digital Omnibus, postponed them in July 2026.
| Date | What applies from this date |
|---|---|
| 1 Aug 2024 | The AI Act entered into force. |
| 2 Feb 2025 | Ban on certain practices and the AI literacy obligation. |
| 2 Aug 2025 | Rules for general-purpose AI models and penalties. |
| 27 Jul 2026 | Changes from Regulation 2026/1744, including a softened AI literacy obligation. |
| 2 Aug 2026 | Transparency obligations, such as telling people they are interacting with an AI system. |
| 2 Dec 2027 | Obligations for high-risk systems under Annex III, including AI in recruitment. |
What is already prohibited
Since 2 February 2025, certain practices have been banned. For HR, the most important is the ban on using AI to recognize people’s emotions in the workplace, except for medical or safety reasons. This covers systems that infer emotions from biometric data, such as the face or voice.
Manipulative techniques, social scoring and biometric categorization that would infer, for example, race, political opinions, trade union membership, religion or sexual orientation from biometric data are also prohibited.
Be careful with interview tools. If a tool estimated a candidate’s emotions from video or voice, check it with a lawyer before using it.
Obligations for companies that use AI in recruitment
The AI Act distinguishes between providers and deployers. A provider develops the system and places it on the market. A deployer is a company that uses the system under its own authority. An employer using recruitment software is usually a deployer.
| Obligation | What it means in practice |
|---|---|
| Use as instructed | You use the system as the provider intended. |
| Human oversight | A person with the right skills and authority checks the AI’s outputs. |
| Relevant input data | The data you feed into the system fits its purpose. |
| Monitoring and reporting | If there is a risk or a serious incident, you inform the provider and the competent authority. |
| Logs | You keep the system’s automatic logs for at least six months. |
| Informing employees | Before using it in the workplace, you inform workers’ representatives and the affected employees. |
| Informing candidates | You tell people that the system is used to help make decisions about them. |
| Impact assessment | Where GDPR requires it, you carry out a data protection impact assessment. |
These obligations apply to high-risk systems from 2 December 2027. The obligation to support AI literacy among people who work with AI already applies. Since July 2026, it no longer requires guaranteeing a specific level of knowledge.
Fines
| Infringement | Maximum fine |
|---|---|
| Prohibited practices | €35 million or 7% of worldwide annual turnover |
| Other obligations, such as those of deployers or transparency | €15 million or 3% of turnover |
| Incorrect or incomplete information to authorities | €7.5 million or 1% of turnover |
The higher of the two amounts applies. For SMEs, the lower amount applies.
How to prepare
- List the AI tools you use in HR. In recruitment, assessment, shift planning and performance monitoring.
- Find out which ones are high-risk. The purpose decides: selecting, assessing, promoting, dismissing or monitoring people.
- Check that you do not use prohibited practices. Especially emotion recognition.
- Ask your vendors. Request documentation and instructions for use.
- Decide who oversees the AI’s outputs. And train them.
- Inform people. Candidates and employees should know where and how you use AI.
- Connect it with GDPR. Handle personal data processing and impact assessments together.
From practice
What to watch out for when using AI in HR, where the legal limits lie and what risks misuse brings were discussed by attorney Daniel Vejsada of PRK Partners in the webinar RECRU Academy #15: AI Act in HR from a lawyer’s perspective (in Slovak).
The AI Act and RECRU
In RECRU, the AI agent suggests suitable candidates, but the recruiter confirms adding a candidate to the database or a hiring project. The voicebot delivers a transcript and assessment of each call, and the recruiter decides on the next steps. When classifying your tools under the AI Act, ask the provider for documentation and instructions for use.
RECRU is a cloud service you can try free for 14 days.
Frequently asked questions
Does the AI Act apply to small companies?
Yes. The obligations depend on what AI a company uses and for what purpose, not on its size. SMEs have lower fine caps and support from national authorities and the European Commission.
Can I use ChatGPT in recruitment?
For writing job ads, emails or interview questions, yes. That is not a high-risk use. If you used it to assess or rank candidates, that is an area the AI Act classes as high-risk. When handling candidates’ personal data, keep GDPR in mind as well.
Is every applicant tracking system high-risk?
Not automatically. It depends on what the AI in the applicant tracking system is used for. If the system ranks, filters or evaluates candidates, it falls into the high-risk category. Article 6(3) sets out exceptions for narrow supporting tasks, which do not apply to profiling.
Does GDPR still apply alongside the AI Act?
Yes. The AI Act does not replace GDPR. When processing candidates’ data, both regulations apply, including the GDPR rules on automated decision-making.
Sources
This text is for information only and does not constitute legal advice. Status as of 18 November 2026.
- Regulation (EU) 2024/1689 on artificial intelligence, EUR-Lex.
- Regulation (EU) 2026/1744 of 8 July 2026 amending the AI Act among others (Digital Omnibus), EUR-Lex, published on 24 July 2026.
Subscribe To Our
HR Newsletter
Stay updated and get the latest scoop on HR topics.
We care about your data in our privacy policy